Although China’s efforts remain relatively unsophisticated, especially compared with Russia’s in 2016, it is noteworthy, analysts say, that Beijing is seeking stealthily to seed its propaganda and disinformation on Western social media platforms.

“While the Chinese Communist Party won’t allow the Chinese people to use Twitter, it is happy to use it covertly to sow propaganda and disinformation internationally,” said Fergus Hanson, director of the International Cyber Policy Center at the Australian Strategic Policy Institute, which issued a report Thursday analyzing the Chinese campaign. “Persistent, covert and deceptive influence operations like this one demonstrate the extent to which the party-state will target external threats to its political power.”

The campaign recently has broadened to exploit racial unrest in the United States, ASPI found. One account, for instance, tweeted an image of Lady Liberty with a knee on the neck of George Floyd, an unarmed black man whose death last month after a white police officer in Minneapolis held a knee to his neck has sparked global protests and calls for police reform.

Twitter’s disclosure that it had removed 23,750 accounts builds on an action last August in which it removed other accounts that the social media company explicitly linked with China’s ruling party.

The accounts were pushing the official, often contested Chinese line on sensitive issues including the pro-democracy protests in Hong Kong (violent radicals encouraged by the United States), the novel coronavirus (Chinese worked together and triumphed over it) and Taiwan (it learned its covid response from China.)

They comprised the highly engaged core of a network that includes some 150,000 “amplifier accounts” that had few or no followers and were strategically designed to artificially inflate metrics to make it appear the tweets were highly popular, Twitter said.

These core accounts fired off 348,608 tweets between January of 2018 and April. Most were in traditional Chinese characters, as well as some special characters used only in Cantonese, which is spoken in Hong Kong. The propaganda campaign was aimed at Hong Kong residents, as well as the Chinese-speaking diaspora, researchers said.

“What we see in this data set is further evidence of a sustained commitment by the Chinese Communist Party to use social media for influence operations,” said Renée DiResta, technical research manager at Stanford Internet Observatory, which also analyzed the data.

Twitter is blocked within China, where censors tightly control the Internet to try to ensure that the government version of events is the only one accessible. But over the past year, China has sought to broaden its online propaganda efforts beyond its domestic audience and has aggressively pushed its narrative in the wider world.

Chinese diplomats – called “wolf warriors” for their pugnacious style – have used Twitter to spread Beijing’s message to international audiences in English and other languages. The government also has begun to surreptitiously target Chinese-speaking audiences around the world using inauthentic or hacked accounts on digital forums such as Twitter and Facebook.

“This large-scale pivot to Western platforms is relatively new, and we should expect continued evolution and improvement” given the government’s resources and interest in promoting its narrative, ASPI wrote.

China’s Foreign Ministry did not answer The Washington Post’s request for comment on the core points of the report. Instead it responded with a general call for countries to work together. “Both the United Nations and the World Health Organization have called on all countries to strengthen unity and cooperation in cracking down on disinformation,” the ministry said in a faxed statement.

Last August, Twitter removed almost 1,000 accounts that it said were linked to the Chinese government and focused on undermining the legitimacy of the Hong Kong pro-democracy movement.

Beijing responded by “immediately creating new [inauthentic] accounts to continue pushing the same themes, which include emergent priorities such as covid, and English language content to surreptitiously push CCP talking points,” DiResta said.

Twitter’s takedown disclosures, she added, and their efforts to allow access to data for researchers, “offer a model for other companies to follow.”

The immaturity of the Chinese campaign is reflected in the low level of authentic user engagement – of real people reading and retweeting or commenting on the posts, and in the speed with which Twitter detected the accounts.

The operators favored speed and scale over quality, ASPI noted. Content appears to have been assembled hastily, with paragraphs squeezed in and images distorted. In some cases the operators didn’t bother to remove spell-check underlines, the think tank said.

Most of the accounts tweeting about covid-19 were created after January, with one large batch emerging on a single day in April. Some tweets pushed the line that China was better at fighting the coronavirus than Taiwan, a self-ruled democracy that Beijing views as a breakaway province. “China is the best anti epidemic country in the world, not Taiwan,” tweeted a since-deleted account using the handle @NicoleS00264634.

The accounts also targeted billionaire Chinese fugitive Guo Wengui, who lives in Manhattan and is close to former Trump adviser Stephen Bannon. Guo, who also goes by Miles Guo and Miles Kwok, previously worked closely with Chinese intelligence officials but is now campaigning to topple the Communist Party. Beijing is seeking Guo’s extradition to face charges including fraud, blackmail and bribery,and once sent security agents to pressure him to cease his accusations of corruption against the party.

Some of the tweets depict Guo as a rat or show images of protesters outside his apartment holding bilingual signs saying “Guo Wengui is a big traitor.” Many focus on Guo’s relationship with Bannon, suggesting that Bannon is happy to be Guo’s mouthpiece for a price.

“The Chinese are so obsessed with their image, anybody criticizing them from abroad gets leveled with thermonuclear strikes,” said James Mulvenon, director of intelligence integration at the U.S. defense contractor SOS International.

Twitter removed the accounts for violations of its policy forbidding platform manipulation, including engaging in deceptive activity or attempts to make accounts appear more popular than they are, and efforts to artificially influence conversations through fake accounts and automation.

“Improving the health of the public conversation is a priority for our company,” Twitter said in a statement. “If we can ever attribute these behaviors to a state-backed information operation, we disclose them to our public archive – the only one of its kind in the industry.”

Twitter traced the accounts to the Chinese Communist Party through specific unblocked IP addresses originating from the country, said a Twitter official, speaking on condition of anonymity to discuss a sensitive topic.

China is several years behind Russia, which has run online disinformation campaigns in Europe and former Eastern bloc countries for years, and notoriously interfered in the 2016 U.S. presidential election with an effort to exploit societal rifts through posts on Twitter and Facebook that reached millions of Americans.

But while Russia seeks to divide societies and devalue the notion of truth, China, analysts say, is focused primarily on delegitimizing dissidents and adversaries and promoting an image of China as a world power.

“It’s still really easy to identify Chinese information operations because the Chinese are mostly interested in refuting criticism and presenting China in the best possible light,” said Mulvenon of SOS International. “The Russians play Bernie Bros against white nationalists, the left against the right and are much more technically sophisticated. The Chinese are not interested in devaluing truth because they think there is a truth to their message.”

Twitter and Facebook began aggressively to take down state-sponsored accounts pushing propaganda and disinformation after the 2016 election. Influence operations from Iran, China, Russia, and several other countries have been detected by Facebook and Twitter in recent years.

Amazon made the announcement in a brief statement on its corporate blog, though it never mentioned the protests or the recent death of George Floyd after a Minneapolis police officer dug his knee into his neck. But Amazon implied recent events drove its decision.

“We’ve advocated that governments should put in place stronger regulations to govern the ethical use of facial recognition technology, and in recent days, Congress appears ready to take on this challenge,” the company wrote. “We hope this one-year moratorium might give Congress enough time to implement appropriate rules, and we stand ready to help if requested.”

Amazon chief executive Jeff Bezos owns The Washington Post.

Facial-recognition technology has emerged as a key battleground for tech giants vying for business from customers eager to use the latest tools of artificial intelligence. The technology can help identify people and is critical for such services as unlocking smartphones and tagging friends in photos on social media. But it’s also taken on a new and controversial life in law enforcement and other areas which have raised privacy and bias concerns.

Privacy critics have criticized Amazon for selling Rekognition to law-enforcement over concerns that it could lead to the wrongful arrest of innocent people who bear only a resemblance to a video image. And studies have shown that facial-recognition systems misidentify people of color more often than white people.

Rekognition is a tool offered by Amazon Web Services, the company’s profitable cloud computing arm, and it’s relatively cheap and easy to use. It can take grainy photos from a security camera or elsewhere and run it against thousands of photos, like a police department’s database of mug shots, to find a potential match.

Oregon’s Washington County Sheriff’s Office, for example, said last year that it spent about $700 to upload its first big haul of photos and paid about $7 a month after that. Amazon, though, hasn’t named the law-enforcement organizations that use Rekognition, or even disclosed how many use the technology.

Amazon is not pulling the product from the market altogether. It will continue to allow organizations that help rescue human trafficking victims and reunite missing children with their families to continue to use its Rekognition technology, it said.

Some of the sharpest criticism of Rekognition came from the American Civil Liberties Union, which on Wednesday said the moratorium didn’t go far enough.

“This surveillance technology’s threat to our civil rights and civil liberties will not disappear in a year,” Nicole Ozer, technology and civil liberties director with the American Civil Liberties Union of Northern California, said in a statement. “Amazon must fully commit to a blanket moratorium on law enforcement use of face recognition until the dangers can be fully addressed, and it must press Congress and legislatures across the country to do the same.”

The ACLU also called on Amazon to stop selling its Ring doorbell cameras. The unit has partnered with hundreds of police forces across the country to provide potential access to homeowners’ camera footage that Ozer says “fuels police abuse.”

The decision to pause use of the technology by police surprised critics such as Jake Laperruque, senior counsel at The Constitution Project at The Project On Government Oversight, an accountability group, because Amazon has attacked reports that found flaws with Rekognition.

“In the past, they’ve been pretty hostile to the very good research that’s been done,” Laperruque said.

He believes the only reason Amazon reversed course on Rekognition is the public outcry over police tactics in the wake of Floyd’s death. Protesters are calling into question a wide range of policing that could lead to racial abuse.

“It’s highlighted a risk of how this technology could be used in its worst way,” Laperruque said.

On Tuesday, IBM said it will get out of the facial recognition business altogether over concerns about how the technology can be used for mass surveillance and racial profiling.

One tech giant that hasn’t announced any changes with the way it sells facial-recognition technology: Microsoft. Two years ago, Microsoft called on the U.S. government to regulate facial-recognition technology, noting that companies aren’t likely to regulate themselves on their own. Microsoft didn’t immediately respond to a request for comment on if it planned changes in the uses of its facial-recognition technology, now that Amazon paused police use of its system.

One of the authors of a study by the M.I.T. Media Lab that found that Rekognition system performed more accurately when assessing lighter-skinned faces called on Microsoft to act.

“Microsoft also needs to take a stand. More importantly our lawmakers need to step up,” the researcher, Joy Buolamwini, said via email. “We cannot rely on self-regulation or hope companies will choose to rein in harmful deployments of the technologies they develop.”

One of the politicians taking up the call to regulate the technology, Sen. Ron Wyden, D-Ore., said facial-recognition systems should not be used without rules to protect Americans against “inaccurate, discriminatory algorithms and misuse.”

“That goes double for people of color, who are more likely to be wrongly identified and subject to FR for no reason,” Wyden said in an emailed statement.

Amazon employees in recent years have pressed the company over similar issues. Two years ago, workers called on Bezos to end the sale of facial-recognition technology to law enforcement agencies and to discontinue partnerships with companies that work with U.S. Immigration and Customs Enforcement.

Amazon has faced criticism in recent weeks regarding race at the company. In April, its top legal executive suggested the company’s senior leaders fend of workplace safety criticism by trying to turn the focus on a black, activist warehouse worker it had recently fired, calling the worker in a leaked memo “not smart, or articulate.” After the comments became public, the lawyer, David Zapolsky said in a statement that his comments were “personal and emotional.”

As protests over the killing of Floyd grew, Bezos took to Instagram in the last week, posting two racially charged letters from Amazon customers complaining over the company’s decision to run “Black Lives Matter” at the top of its e-commerce site. Bezos wrote that he’s received “a number of sickening but not surprising responses” to his posts. And he said those were “the kind of customers I’m happy to lose.”

“This is my last hurrah,” said Alshemaimry, 70. “I’m going to do this and then retire.”

A year ago his company, OceanBased Perpetual Energy, agreed to work with Florida Atlantic University to develop a way to generate electricity by harnessing the steady-flowing Gulf Stream, the powerful ocean current that brings warm water from the Gulf of Mexico into the Atlantic and up the East Coast to Canada. Now his company was ready for the first test of five types of turbines to see which one would work best while anchored 80 feet below the ocean’s surface.

A successful test, Alshemaimry said, would lead to a project that would cost an estimated $16 billion. The goal: in five years, producing 5 gigawatts of electricity from turbines spun by the Gulf Stream, which would be sent through underwater cables to a power distribution station built in the West Palm Beach area.

The 12- person team submerged the turbines in the Gulf Stream current approximately 20 miles offshore between Broward and Palm Beach counties, where the current’s velocity typically ranges from 3.5 to 5.5 miles per hour. They left them there for 24 hours to see which ones would spin the best in the Gulf Stream’s flow, producing power with the fewest problems.

On June 1, the company declared the test to be a big success, with “game-changing implications for the future of perpetual clean energy.” All of the turbines worked well, but the team selected a design that looks like a pair of airplane engines mounted on a single wing to eliminate the torque caused by the rotating propellers.

Ocean energy works very much like wind power – the force of the sea turns the propellers of a turbine, activating a generator to produce electricity. But submerged turbines come with unique challenges – electrical parts have to be sealed and must resist corrosion, while underwater repairs are disruptive and difficult.

Producing energy from the ocean is not a new idea. The La Rance tidal power station in Brittany, France, has been using 24 turbines to convert ocean tides into electrical power since 1966. Ocean power produces none of the carbon emissions linked to climate change, and it appeals to some energy executives because tides and currents are predictable, unlike solar and wind. But the cost of building the complex infrastructure required is so great that, so far, solar and wind have outpaced it. For instance, the world’s biggest tidal power station, located at Sihwa Lake in South Korea, cost about $300 million to build in 2011.

“It’s safe to say that ocean energy is not a sizable focus in the U.S. currently,” said Fred Mayes, senior energy analyst with the U.S. Energy Information Administration.

A tidal energy project off the Oregon coast, Ocean Power Technologies’ Reedsport Wave Park, fell apart in 2014 when the company was unable to line up financing. The company had obtained a Federal Energy Regulatory Commission permit – the only one ever issued in the U.S. for a wave energy project – to install 10 buoys, each the size of a bus, that would generate enough power for 1,000 homes. The company never deployed any buoys and abandoned the permit when it could not raise the $50 million in estimated construction costs.

“This is a disappointment on many fronts,” Jason Busch, executive director of the nonprofit Oregon Wave Energy Trust, told The Columbian in Vancouver, Wash., at the time. “It has become, rightly or wrongly, sort of a proxy for the industry.”

There is one other American project underway. With state and federal grants, Verdant Power has installed turbines underwater in the east channel of the East River in New York as part of a demonstration. However, the power being generated is below the 1 megawatt threshold required for reporting outputs to the Energy Information Administration.

“In general, we have found that many of these niche applications, while interesting and helpful for research purposes, can’t compete in the wholesale power market,” said the agency’s Glenn McGrath. “However, we continue to monitor developments in this area.”

Gabriel Alsenas, director of the Southeast National Marine Renewable Energy Center at Florida Atlantic, said that’s in part because ocean energy hasn’t been given the same government subsidies that solar and wind industries have enjoyed. As a result, he said, the technology is “immature.”

At least one study says it shows promise, however, A 2013 study of ocean current energy resources by the Georgia Institute of Technology found that the Florida Current portion of the Gulf Stream, which runs like an underwater river along the entire Eastern Seaboard, could generate between 4 and 6 gigawatts. One gigawatt, which equals a billion watts, would be enough to power about 725,000 homes.

Alshemaimry, a Saudi entrepreneur with prior experience building solar-powered homes, spent several years working on a never-completed tidal energy project in Sweden. Then, at a conference in Scotland, he met a U.S. Department of Energy official who suggested he contact Alsenas at Florida Atlantic University about the use of ocean currents.

FAU’s Southeast National Marine Renewable Energy Center has spent more than a decade researching issues related to ocean energy, such as how fish and other marine creatures avoid underwater turbines, and which types of turbines work best in different conditions.

After one phone conversation three years ago, Alenas said, Alshemaimry dropped his Swedish project, switched from waves to currents and moved his entire operation to Florida. Alshemaimry, now contends what he has planned will be far superior to all tidal energy plants.

“Tidal is not 24/7 power,” he said. “It’s back and forth. . . . The Gulf Stream flows 24/7/365.” He calls it “the Holy Grail of perpetual energy.”

While it’s true the Gulf Stream flows all day and night, it’s not perfectly constant: The 2013 study found that its flow varies, growing stronger in the fall and weaker in the spring. Climate change projections say its flow is likely to weaken further and to switch its path someday.

But Alshemaimry said it would continue flowing sufficiently to keep on driving a powertrain, generating as much electricity as Florida could need. That’s a tall order: Florida currently consumes 220.7 terawatt hours of power, most of it provided by plants burning coal or natural gas.

Asked why, if the Gulf Stream is such a perfect resource for power, no one else has tried it before, he said, “No one else is crazy like me.”

Testing of the first two Avelia Liberty high-speed train sets from French manufacturer Alstom, is underway in the Northeast Corridor and at a federal facility in Pueblo, Colorado, and Amtrak said railroad crews have started training on the new technology in anticipation of a launch next spring.

The Acela prototype arrived at the Colorado Federal Railroad Administration site for nine months of testing in February. Officials said it recently exceeded performance expectations, traveling at 165 mph, above the 160 mph limit the trains would be allowed to travel once in service between Washington and Boston. The current Acela trains travel 150 mph.

“We are laser-focused on delivering this new fleet of trains,” said Caroline Decker, Amtrak’s vice president for the Northeast Corridor. “Looking at where we are in terms of the production, we have a high degree of confidence that a 2021 launch is very doable, and certainly we’re eager to introduce the new fleet to the Northeast Corridor as soon as possible.”

The $2.5 billion project, which also includes major infrastructure improvements to accommodate the new trains, is moving forward at a time when Amtrak is preparing to reduce staff by up to 20 percent and is requesting nearly $1.5 billion more in federal aid to keep afloat amid the unprecedented financial hardship from the pandemic. The health crisis that shut down much of the country in March devastated the passenger railroad’s ridership and revenue.

The investment in the Acela was originally aimed to grow one of Amtrak’s strongest lines. Post-pandemic, the investment offers hope for the future as the company tries to recoup from massive losses suffered when ridership plummeted with the health crisis. Railroad officials say they are determined to keep the project on track, saying it could stimulate economic recovery.

The contract for the 28 trains was awarded in 2016 and supports about 1,300 jobs across the country, officials said, including 400 at Alstom’s facilities in areas of support such as train control, rail signaling, engineering and maintenance. The new trains will replace the existing fleet of 20 sets starting next year.

“It’s a silver lining at a time of a lot of dark clouds,” Decker said. “We’re very mindful that we are in very tumultuous times, but I will say this keeps us very motivated, very focused on what is going to be a real game-changer for train travel.”

The entire new fleet should be in operation in 2022, when Amtrak hopes demand for train travel will have rebounded to pre-pandemic levels. Acela, Amtrak’s premier service, was performing at an all-time high before the crisis. Ridership on the Acela grew by 4.3 percent in fiscal year 2019, compared with the previous year, and at a higher pace than the growth on the Northeast Regional and the company’s state-funded routes. Acela’s revenue also grew by about 5.4 percent, according to Amtrak.

That success led Amtrak to pursue expansion of the service, adding nonstop trips between Washington and New York last fall and an additional Washington to New York to Boston round trip on Saturdays.

The new trips showed promise, officials said, until the coronavirus hit and the Acela, popular for business travel, was among the first services to be cut as the virus began to spread and demand for travel sank. Some Acela trips resumed Monday, but officials said the railroad doesn’t expect it or its entire network of intercity passenger trains to return to normal anytime soon. It is even less clear when, or if, the new Acela nonstop will return.

In late May, Amtrak chief executive William J. Flynn said the company was projecting a 50 percent reduction in systemwide revenue in fiscal 2021, saying demand remains about 5 percent of normal. The company estimates ridership in the next fiscal year may reach 16 million, or roughly 50 percent of the pre-pandemic levels.

– – –

The new trains are being built with several touchless and self-serve features that Amtrak says should make train travel more appealing in the post-coronavirus era with Americans still fearful of infection spreading through communal surfaces and human contact.

The lavatories are more spacious, accessibility compliant and have touchless and contactless door and faucets. The cafe car will have self-select and self-serve options.

Once the trains go into service, Amtrak plans to implement reserved seating, which the company says may help reduce the long and often tumultuous lines at the station as passengers rush to board all at once to grab a seat.

Among other features: additional interior and exterior signage to assist passengers in finding their way, streamlined overhead luggage compartments and doorless luggage space so passengers have fewer surfaces to touch. Power outlets and USB ports are more accessible to both passengers in between the seats.

Amtrak officials said it is working with the manufacturer to reevaluate the interior design to determine if any other enhancements can be made in response to the pandemic.

“Are there additional features that we could incorporate to provide better and more enhanced safety for the traveling public?” Decker said. “We’re going to do everything we can to continue to improve the product.”

In addition to their faster speed, the trains can accommodate up to 386 passengers, an increase of 25 percent.

Earlier this year, Amtrak’s inspector general warned that the company’s plan to roll out the trains early next year could be derailed, citing delays in their delivery, testing and training. Infrastructure improvements, including modifications to three maintenance facilities needed to get the trains into service also were behind schedule, according to the inspector general’s report.

Amtrak said last week that it awarded contracts, and work is underway for modification of the maintenance facilities in Washington, New York and Boston. It also said a team of Amtrak engineers is working closely with the manufacturer. Various work groups are leading training, testing and other preparations.

The original plan to roll out the trains starting in January had already been pushed back a few months, Decker said. The expectation now, she said, is to have the first trains enter service in late spring or early summer next year with nine of the 28 train sets operating by fall 2021. That, however, is dependent not only upon the manufacturer completing the train sets but also successful completion of rigorous testing and training.

As of this week, the testing on the first two trains was progressing, Amtrak said. The high-speed testing of the train set based in Colorado completed a “milestone” last month when it traveled at speeds up to 165 mph. That train has six months of testing before it returns to Hornell, New York, for installation of interiors.

On Monday, the second prototype, based for testing in Philadelphia, made its way to Washington. The train is undergoing testing on the same tracks where Amtrak passengers are carried and will be in the corridor through the end of the year. Testing evaluates the trains’s performances and safety, from railway dynamics to traction, brakes and train control management systems. After a successful testing, the train also will return to Hornell to be completed.

“The project is on track,” Alstom spokesman Michelle Stein said, confirming that the 28 trains “will be produced and delivered by 2022.”

Production at the Alstom facility in Hornel slowed in March as New York shut down in its effort to halt the spread of the novel coronavirus. The work at the factory, however, was deemed essential, and although there was a reduction in personnel, it didn’t close completely.

Alstom said the impacts have been manageable and minimized in part because of a “strong supply chain of nearly 250 suppliers in 27 states” and because 95 percent of the components for the new train sets are produced domestically. For the testing, Stein said, Alstom deployed advanced digital remote monitoring capabilities, which have allowed the testing program to continue in Colorado despite travels limitations.

Some production activities continued during the crisis, such as component testing, cabling and wiring work, and other warehouse jobs that could be done with social distancing, Amtrak and Alstom said. As of this week, Alstom had recalled all production workers back to the site and work was ramping up, officials said.

As the testing of the first two trains gets underway, in New York more than 250 workers are working on the production of two additional trains at the Hornell factory.

As Amtrak works toward recovery from the pandemic, the new trains are “essential to improving reliability, service, safety and capacity” for travel along the Northeast, Sen. Charles Schumer, D-N.Y., said in a statement. Earlier this year, Schumer urged the Federal Railroad Administration to begin the federal inspection and testing of the train sent to the Colorado facility, saying that keeping the project on schedule was critical.

“The Next Generation Acela train sets, built by our world-class workforce in Upstate New York, will help Amtrak grow revenue and ridership for years to come,” Schumer said this week.

So two months after his school closed, and with next school year on the horizon, Herr, a sixth-grade science teacher in Washington, D.C., wonders: Can he go back when classrooms reopen? Will he be safe? How will he know?

School districts across the country are sharing rough drafts of what the fall could look like. They are under increasing pressure, from parents and politicians, for those plans to include at least some in-person learning.

But teachers, especially ones who are older or medically compromised, worry those plans do to little to protect them.

The plans are also just unrealistic, teachers say. They can’t envision students maintaining social distance, keeping masks on, or walking in the same directions in hallways, all things health officials are recommending. Even before the pandemic, teachers said, their schools struggled to keep ample soap and water running in the bathrooms.

“When I hear about keeping students socially distant, I just kind of laugh at that,” said Crysta Weitekamp, a 47-year-old special education teacher at Southeast High School in Springfield, Ill., who has asthma. “They’re social creatures.”

So, teachers say, they’re anxious about returning. But they’re also anxious about what happens to their job if they refuse.

“It does make me nervous to say no,” said Lara, a high school teacher at a Los Angeles charter school who is immunocompromised. (Worried about her job, she spoke on the condition that she be identified only by her first name.) “Of course schools need to reopen, but at what point are you being sacrificed?”

For all the plans to reopen schools, from masked kids to staggered schedules to half-empy buses, few address what to do with at-risk teachers. But one strong advocate of a return to normalcy this fall has at least brought it up: President Donald Trump said that schools “should be opened ASAP,” and that older teachers may just need to stay home.

The American Enterprise Institute, a right-leaning think tank, has an idea for making that happen: It recommends school districts offer early retirement incentives to older teachers, a controversial proposal that has long been pushed by some education reformers – and dismissed by teachers unions as a way to drive down labor costs.

John Bailey, co-author of the American Enterprise study, said this is an unprecedented scenario and that the plan would protect older teachers. Given the economy, he is unsure how many teachers would even want this option; districts should also find new, less risky roles for those teachers. But retirement incentives are worth exploring, he said.

Nearly 18 percent of teachers are 55 and older, according to the National Center for Education Statistics. If teachers retire in big numbers, that could exacerbate a national teacher shortage, particularly in rural areas.

“They are being squeezed on both ends,” Bailey said of school districts. “They are having these teachers who cannot come to school. And they are also having to find teachers who are able to come to school.”

Randi Weingarten, president of the American Federation of Teachers, the nation’s second largest teachers union, told teachers to “scream bloody murder” if their districts attempt to reopen against health guidelines. But as pressure builds for an in-person fall, teachers and unions say they want to be part of the discussion so they can ensure that reopening is driven by science, not economics.

“There is a lot of categories of people who we should be concerned about in the absence of a vaccine and in the absence of very aggressive testing,” Weingarten said. “The notion that it’s just affecting people 65 and older is wrong.”

School leaders across the country are proposing alternatives. One Idaho school that has already reopened is allowing teachers to lead classes remotely from home, with substitute teachers supervising the physical classrooms, according to Education Week.

In Washington, Paul Kihn, deputy mayor for education, called figuring out how to staff school in the fall a “jigsaw puzzle.” The pieces of that puzzle include teachers who are considered high-risk; teachers who live with someone who is high-risk; and teachers who have children on a different school schedule. City health guidelines call for teachers who are considered high-risk to receive medical clearance before returning to the classrooms.

“This is an extraordinarily complex undertaking for each school community,” Kihn said. “We have a fairly good understanding of the concerns that families, parents and school staff have about returning to school buildings. We also understand that our primary job is to ensure safety.”

Kihn said the school district is exploring plans to protect at-risk teachers. One factor, he said, is that many students won’t attend in-person classes because they have a medical condition or live with someone considered high-risk. Teachers who must remain home may be able to teach those children virtually. And like many districts, D.C. is considering having students report to school for in-person learning some days and stay home on others, to allow for more social distancing on campuses. At-risk teachers may be given a greater share of distance-learning.

Eric Williams, superintendent of Loudoun County Schools and a member of Virginia Gov. Ralph Northam’s, D, Education Work Group, said reopening can happen only if school districts are entirely transparent with their teachers and staff. Opening up won’t happen unless there are good plans in place “so staff and parents won’t have anxiety about returning,” he said. “We are committing to making decisions based on conditions, not time. It’s about data, not dates.”

With so many unknowns about how long the pandemic will extend, teachers – including younger and healthy ones – question whether schools can safely pull off a plan.

Herr’s 33 year-old wife, for example, is a librarian at a D.C. Public Schools campus. She is healthy, he said, but she worries her daily interactions with children could endanger Herr.

“These are kids, and they are unpredictable, and we can set up as many structures and guidelines as we want, and stuff is going to fall through the cracks, and kids will not always follow the rules,” Herr said. “The consequences of that are really dire.”

Teachers say the safe reopening of schools relies on families following the health guidance. If someone in a household is ill, school staff said they must trust that the family won’t send their kids to school. But job pressure often forces parents to send their kids to school sick, and teachers said they don’t believe that will change.

“I want to be back in school with students, badly,” said Zach Carroll, a middle school teacher in the District. “But would I feel safe? Not at all.”

Teachers aren’t the only school-based staff worrying about returning. Paraprofessionals, cafeteria workers and security guards tend to skew older. In D.C. Public Schools, 2 percent of teachers are 65 and older, according to city data. But 6 percent of school-based staff are 65 and older.

Robert Alston Jr., the in-school suspension coordinator at Coolidge High in Washington, is the leader of the union that represents school support staff. He’s 51 and has type 2 diabetes. When a fight breaks out at his school, he’s the person called to help break it up. When a teenager at his school is struggling or has an outburst, he often rubs their backs, soothes them, ensures they are OK. He wonders how aides assigned to help students with special education needs would continue their work under these health guidelines.

“We talk about front-line workers, but do they really look at us – support staff in schools?” Alston asked. “We have people who are crying, kids who are bleeding, kids who need a hug, kids who need extra support.”

Alston said support staff are in this profession for the long run. So whatever schools plan for the fall, they need to figure how to accommodate him and his colleagues.

“We are the lowest-paid people in schools,” Alston said. “We don’t have enough money to retire.”

Trump last week tweeted that “when the looting begins, the shooting begins,” which many people interpreted as a call for violence in nationwide protests over the death of George Floyd. Twitter put a warning label over the tweet, flagging it as violent content that broke the company’s policies but was being left up because it was newsworthy. Facebook declined to take any action on a similar post on its site.

In response, dozens of Facebook employees participated in a virtual walkout on Monday, and many more expressed outrage in internal forums and on Twitter. At least two employees have resigned, according to public posts and tweets and conversations with workers.

“Open and honest discussion has always been a part of Facebook’s culture,” spokesperson Liz Bourgeois said in a statement. “Mark had an open discussion with employees today, as he has regularly over the years. He’s grateful for their feedback.”

During the town hall, Zuckerberg did not back down from his decision to keep up the post, as many employees had hoped, according to several workers who were listening but declined to provide their names for fear of retribution.

At least five people have died in nationwide protests that began this weekend. Thousands more have been tear-gassed and injured.

Facebook’s policy says it removes language that incites or facilitates serious violence.

At the town hall, Zuckerberg defended his decision that the post did not constitute a policy violation, as he personally walked employees through different interpretations of Trump’s language.

But the CEO said he would begin to review the transparency of the processes around how pieces of content get escalated to senior managers. He also said that he would be open to reviewing how the company handles content around state violence, a nod to the growing use of force at the protests.

Zuckerberg’s personal involvement in the decision is characteristic of the way he has handled controversial policy choices over the last several years, since the social network has been in a state of crisis over accusations of Russian meddling and the Cambridge Analytica privacy scandal. His leadership style contrasts with Twitter CEO Jack Dorsey, who tends to delegate policy decisions to his deputies.

Zuckerberg also made the decision not to take down a video of House Speaker Nancy Pelosi that was manipulated to make her appear drunk. He made a personal call not to fact check political advertising, despite frustration from the public and from employees, according to a person familiar with the decision-making.

Historically, both Facebook and Twitter have been hands off when it came to political speech, exempting it from policy violations because it is newsworthy. But Twitter has recently pivoted from its long-held stance. The company has decided to fact check politicians and to label their commentary with a warning when it breaks the service’s policies, as it did for the first time with Trump last week. Twitter first fact-checked a Trump tweet with misleading information about mail-in ballots.

Twitter’s decision was celebrated in liberal Silicon Valley.

Facebook, on the other hand, has doubled down in the direction of free speech, with Zuckerberg reiterating the point in speeches and public statements last year.

Two of the people who attended the town hall said Zuckerberg’s suggestions seemed like minor concessions that did not appear to appease the many angry employees, some of whom repeatedly pointed out in questions that very few black people were attending the town hall.

The decisions at Facebook have prompted at least two employees to publicly resign.

Timothy Aveni, a software engineer according to his Facebook page, said in a public resignation letter that he was disappointed in Zuckerberg’s leadership.

“Mark always told us that he would draw the line at speech that calls for violence,” Aveni wrote. “He showed us on Friday that this was a lie. Facebook will keep moving the goalposts every time Trump escalates, finding excuse after excuse not to act on increasingly dangerous rhetoric.”

Twitter’s decision to flag two of Trump’s erroneous tweets last week for the first time prompted the president to lash back, signing an executive order that called for reexamining a law that has helped shield tech giants from liability for content posted on their sites.

On Tuesday, Washington-based advocacy group Center for Democracy and Technology, which is supported by Facebook, Google and Twitter, filed a lawsuit alleging the executive order threatens to “curtail and chill constitutionally protected speech” across the web.

Throughout the weekend, Facebook employees used the company’s internal chat system to question not only Trump’s tweets but a handful of others tweets by politicians that appeared to encourage violence, including by Florida GOP Congressman Matt Gaetz, which Twitter also labeled.

Electronic Arts also scrapped an event to introduce the Madden NFL 21 game that was set for Monday. Airbnb said Chief Executive Officer Brian Chesky won’t deliver a planned video message to discuss the home-rental startup’s vision for travel. And Alphabet’s Google postponed the introduction of its Android 11 mobile operating system previously planned for June 3.

Demonstrations against the killing of an unarmed black man, George Floyd, by a white police officer in Minnesota last week have turned violent in cities from New York to Los Angeles. Officials have set curfews in major cities to deter late-night protests and looting. The situation has reopened racial wounds and cast a somber tone in the country.

For the PS5 event, which had been scheduled for June 4, Sony said, “We do not feel that right now is a time for celebration, and for now, we want to stand back and allow more important voices to be heard.”

Electronic Arts issued a statement with a black background that said: “We stand with our African American/Black community of friends, colleagues and partners.” The company said, “We’ll find another time to talk football with you because this is bigger than a game, bigger than sports and needs all of us to stand together and commit to change.”

After the president tweeted a message with the words “when the looting starts, the shooting starts” in response to protests over the death of George Floyd in Minneapolis, Twitter for the first time obscured one of his posts, marking it with a warning that it breached service rules by glorifying violence. Facebook’s response to the same content, in a post from Zuckerberg on Friday, was to say, “We think people need to know if the government is planning to deploy force.”

Several senior figures at Facebook expressed strong disagreement.

“Mark is wrong, and I will endeavor in the loudest possible way to change his mind,” said Ryan Freitas, director of product design for Facebook’s News Feed. “I apologize if you were waiting for me to have some sort of external opinion. I focused on organizing 50+ likeminded folks into something that looks like internal change.”

“Giving a platform to incite violence and spread disinformation is unacceptable, regardless who you are or if it’s newsworthy,” wrote Andrew Crow, head of design for Facebook’s Portal product line.

Joining them with individual messages against the passive policy were Design Manager Jason Stirman, Director of Product Management Jason Toff and Product Designer Sara Zhang, who tweeted that “Internally we are voicing our concerns, so far to no avail.”

Read more: Facebook Appeases Trump as Twitter Spars With Him Over Posts

In a post late Sunday, Zuckerberg said Facebook is committing “an additional $10 million to groups working on racial justice.” Noting that the company “has more work to do to keep people safe and ensure our systems don’t amplify bias,” the CEO did not address the concern surrounding Trump’s posts on the platform.

It’s rare for Facebook employees to speak publicly about internal activity unless they have permission from the communications team. The company in the past has punished and discouraged leaking. Now, Facebook has changed that approach.

“We recognize the pain many of our people are feeling right now, especially our Black community. We encourage employees to speak openly when they disagree with leadership,” a Facebook spokesperson said Monday in a statement. “As we face additional difficult decisions around content ahead, we’ll continue seeking their honest feedback.”

Before that happens, the astronauts and controllers on the ground have to ensure that the pressure between the spacecraft and the station is equalized. They will also be setting up an umbilical that will allow communications and power to transfer between the two.

The Dragon spacecraft, now named Endeavour, docked at 10:16 a.m. EDT, a few minutes earlier than planned. Shortly after docking, Hurley said, “it’s been a real honor to be just a small part of this nine-year endeavor since the last time a United States space ship docked with the International Space Station.”

In Houston’s mission control, flight director Zeb Scoville congratulated the crew.

“Bravo on a magnificent moment in spaceflight history,” he said, “and on the start of a new journey that has changed the face of space travel in this new era of space transportation.”

Before opening the hatch and entering the station, Behnken and Hurley will conduct a series of pressure and leak checks to ensure their safety. Then they will join fellow NASA astronaut Chris Cassidy and two Russian cosmonauts aboard the station.

The docking was a delicate and dangerous part of the mission. The spacecraft chased down the space station, traveling in orbit at 17,500 m.p.h., but then approached very slowly in a series of carefully choreographed maneuvers.

The mission went smoothly, ground officials said, following a picture-perfect launch some 19 hours earlier from the Kennedy Space Center.

The flight is the first launch of NASA astronauts since the Space Shuttle retired in 2011, and the first of a private company of humans to orbit.

Behnken and Hurley said during a live broadcast Sunday morning that they have had a smooth ride so far, got a decent night’s sleep and awoke this morning to Black Sabbath’s “Planet Caravan,” continuing a NASA tradition of waking astronauts to music.

En route to the space station, the Dragon spacecraft performed a series of “burns,” or engine thrusts that raised its orbit to eventually match that of the station. Shortly before 9 a.m. the spacecraft was moving into position about 400 meters below the station.

The Dragon spacecraft flies autonomously, but the astronauts can take over the controls at any time, and seem to be enjoying flying a modern spacecraft. During a broadcast from the capsule, Hurley noted that they were the first astronauts to control a spacecraft using a touchscreen.

“So we got that going for us,” he said.

But its exterior isn’t nearly as intimidating as its appetite. By the time a crown-of-thorns matures at 4 months old, it eats live coral voraciously – up to its own body weight in one evening. Just one can consume 20 to 32 feet of living coral a year. Thirty of them per 2 1/2 acres can kill an entire reef’s coral.

These starfish are the second biggest threat to the already endangered Great Barrier Reef – the world’s largest coral reef system, right behind tropical cyclones. Over the past decades, they have attacked the reefs in waves of outbreaks, one of which spiked their population to as much as 1,000 starfish per 2 1/2 acres.

The Great Barrier Reef covers nearly 865 million acres off the coast of Australia, which means approximately 350 billion starfish inhabit it. The starfish, the most fertile invertebrate in the world, stripped 150 reefs of coral within the Great Barrier system and damaged 500 more in just a few years.

These enormous outbreaks initially confounded scientists, but they have begun to make progress in understanding them – discovering a surprising biological trait – and in fighting them with the development of an underwater robot to attack them.

The recent discovery that crown-of-thorns starfish, called COTS for short, can delay their transition to adulthood for at least six years means they can lie in wait without needing to feed on coral while a damaged reef heals itself, then feast on it as soon as its healthy coral has grown back.

“Because COTS juveniles have the ability to stay in an algae feeding form for up to six years, there could be an accumulation of multiple generations of juveniles that are happily feeding on algae until there is a specific cue that catalyzes their transition to feeding on corals,” says Paul Barber, professor in the Department of Ecology and Evolutionary Biology at the University of California at Los Angeles.

Imagine hundreds of thousands of crown-of-thorns living that way around a single reef system – its living coral couldn’t withstand the assault. This starfish’s development suspension ability isn’t the only driver of outbreaks. Human interference is responsible as well.

“The science tells us that elevated fertilizer and other pollutants from primary production runoff is causing an increase in phytoplankton, which is the main food source for COTS larvae,” says Anna Marsden, managing director of the Great Barrier Reef Foundation.

Overfishing and the removal of the starfish’s natural predators, like the giant triton snail, have also contributed to outbreaks.

Under normal circumstances, crown-of-thorns starfish play a useful role in reef vitality and diversity because they prefer to eat the faster growing coral, which helps slow-growing coral varieties gain a foothold. However, when their population isn’t kept in check, they can easily become a devastating invasive species.

The Great Barrier Reef Foundation has been working with local farmers to help reduce the amount of pollutant runoff into the reef’s waters, which will hopefully lessen the number of outbreaks. It recently launched a $37.5 million crown-of-thorns starfish control program that is exploring innovative surveillance and control methods.

“The program focuses on priority reefs, defined based on their ecological and economic significance,” Marsden says.

Scientists have been exploring methods to mitigate crown-of-thorns outbreaks since the early 1960s, when the escalation in outbreaks was first noted. In 2015, researchers at James Cook University discovered that a 20 milliliter dose of vinegar would kill a starfish in 48 hours. Before that, the only known injectable poison was bile salts, which are expensive and harder to get in large amounts.

But coming up with a good culling tool was only half the battle. To administer either of these injections, diving crews had to hunt down the starfish one by one. As a result, it’s been difficult to make much of a dent in their population.

For example, in 2015 the Great Barrier Reef was home to anywhere between 4 and 12 million crown-of-thorns starfish. In one year, two crews working full-time were only able to eradicate 350,000. Each female lays approximately 65 million eggs a year. Clearly, crown-of-thorns starfish control via human divers alone is an uphill battle.

Enter the RangerBot. The RangerBot is an Autonomous Underwater Vehicle (AUV) and the first vision-based robot designed to protect coral reef systems. It was developed by a team of scientists led by Matthew Dunbabin, professor of science and engineering at Queensland University of Technology over the past six years. “Given the extreme size of the Great Barrier Reef and many threats to humans, we wanted a ‘tool’ that could allow the authorities to scale back the manual eradication program,” Dunbabin says.

The idea for the bot was born 14 years ago. “In 2005, I developed a robot which proved that vision can be used to allow a robot to estimate its position in the reef and avoid obstacles,” Dunbabin says.

However, computer technology was not yet advanced enough to allow for real-time detection of crown-of-thorns. That changed in 2014 with advancements in Deep Learning technology, which allowed Dunbabin’s team to construct a robot that could be programmed to detect and inject the starfish with bile salts all while operating autonomously. Over the next few years, they refined the prototype to be smaller, lighter and less expensive so that a bot army could eventually be commissioned.

In 2016, the RangerBot won the Google Impact Challenge Australia and joined forces with the Great Barrier Reef Foundation in an effort to accelerate the development process. By late 2018, the RangerBot was fully operational in the Great Barrier Reef. In 2019, the team was able to increase its capabilities so that it can also help with coral larvae reseeding along the Great Barrier Reef and reefs in the Philippines.

“(We) have been integrating them with robotic boats to further increase their ability to deliver large amounts of coral larvae to damaged reefs,” Dunbabin says.

The endeavor has not been without challenges, though.

On the technological side, they’ve had issues getting the bot’s vision algorithms to run in real-time once it’s on an underwater mission, so it can be difficult to precisely track it. But the bigger challenge is the evasiveness of the crown-of-thorns starfish themselves.

When they’re all gathered together, the RangerBots have an advantage, but during the day the crown-of-thorns tend to hide under coral, which makes it harder for a bot to see and inject them. “There are some studies that show that COTS come out more to feed at night, so our plan has been to use the lights on RangerBot to do nighttime operations and possibly increase its utility further,” Dunbabin says.

The Great Barrier Reef is an irreplaceable wonder of the natural world. Its 3,000 reef systems spanning over 214,000 miles – roughly the size of New Zealand – are home to more than 1,500 species of fish, 400 types of hard coral and one-third of the world’s soft coral. In just the past 30 years, half of that coral has been lost, largely because of bleaching events and crown-of-thorns outbreaks.

Scientists like Dunbabin hope research will find a way to keep the starfish plagues in check. His effort, still undergoing study and troubleshooting, has aroused the interest of conservation groups considering supporting production of RangerBot armies. That, he says, could happen within the next 12 months.