Brandon Roberts, Propublica

What Happened: A trio of lawmakers demanded transparency from the Department of Veterans Affairs on Tuesday, saying the Trump administration continues to “stonewall” requests for details on the agency’s recent cancellation of hundreds of service contracts.

The group, which included Sens. Richard Blumenthal and Angus King, as well as Rep. Mark Takano, said that despite repeated requests, the agency has disclosed incomplete and inaccurate lists that failed to specify exactly which contracts have been canceled. Blumenthal and Takano are Democrats, and King is an independent. They made their comments at a special forum in Washington.

A review by the Democratic members of the Senate Committee on Veterans’ Affairs identified 655 contracts canceled by the VA, where previous lists disclosed by the agency included dozens less and contained significant errors.

The lawmakers cited a recent ProPublica investigation into the agency’s use of a flawed artificial intelligence tool to assess VA contracts. That analysis was conducted by a staffer from the Department of Government Efficiency with no health care or government experience. The VA uses contractors for a range of services, including to support hospitals, research and other services aimed at caring for ailing veterans.

What They Said: Lists of contracts previously disclosed to the committee are “gobbledygook” and filled with errors, the lawmakers said. “This hearing shouldn’t even be necessary,” said King, who sits on the VA oversight committee. “The simplest thing is to send us a list.”

Senators highlighted the harm caused by canceling the contracts, including one that resolved glitches between VA systems preventing veterans from receiving benefits. Without this contract, said Benjamin Ambrose, whose job it was to resolve these errors, there is nobody left at VA to do this work. “In this case veterans are being locked out forever,” he said.

Scott Amey, general counsel with the bipartisan Project on Government Oversight, said: “There’s a lot of fallout. There’s a lot of dominoes that go with canceling just one contract.”

Amey expressed doubt that the necessary work was done to ensure canceled contracts were duplicative or wasteful. “From the stonewalling that we’ve heard from the VA, you can’t have any confidence that that work was done,” he said.

The lawmakers also questioned the VA’s use of AI to assess contracts for possible cancellation, referring to ProPublica’s investigation. Blumenthal said AI holds promise, but it “has to be used thoughtfully.”

Background: ProPublica reported on Friday that the VA used an error-prone AI tool to identify contracts for possible cancellation. The tool, written by former DOGE staffer Sahil Lavingia, used outdated AI models to “munch” contracts based on conflicting instructions and produced glaring mistakes, a ProPublica analysis found.

Experts in AI and government procurement agreed that the DOGE analysis of VA contracts was flawed, with one calling it “deeply problematic.” Lavingia acknowledged that there were problems. “I’m sure mistakes were made. Mistakes are always made. I would never recommend someone run my code and do what it says. It’s like that ‘Office’ episode where Steve Carell drives into the lake because Google Maps says drive into the lake. Do not drive into the lake.”

ProPublica identified at least two dozen contracts on DOGE’s list that have been canceled so far. Among them is a service agreement to maintain a gene sequencing device used to develop better cancer treatments. Another was with Columbia University for blood sample analysis to support a VA research project. Others still were related to addressing nursing issues, including one to develop social media tools to recruit nursing staff and another to help assess and improve the care they provide.

Democrats in Congress have been seeking more information from the VA on the canceled contracts in an attempt to assess whether the cuts have put veterans’ well-being in jeopardy.

Response: VA press secretary Pete Kasperowicz has defended DOGE’s work on reviewing contracts, saying that the vetting sets a “commonsense precedent.” He and Lavingia have said that VA staffers reviewed everything on the DOGE “munchable” list before deciding which contracts to cut.

In a statement on Tuesday, Kasperowicz said that the agency’s contract review has been a careful process aimed at benefiting veterans and using taxpayer money efficiently. “Decisions to keep, cut or descope contracts are based on careful and methodical multilevel reviews by VA employees, including career subject-matter experts who are responsible for the contracts, as well as VA senior leaders and contracting officials,” he said.

He disputed any suggestion from legislators that the contract review might diminish essential services. “Terminating or not renewing these contracts will not negatively affect veteran care, benefits or services,” he said. “In fact, these decisions will allow VA to redirect billions of dollars back toward health care, benefits and services for VA beneficiaries.”

Why It Matters: Over 9 million veterans across the U.S. rely on the VA for health care through its network of 170 hospitals and 1,200 clinics. One of the nation’s largest health care providers, it is a training ground for doctors and nurses and an engine for medical research. Since returning to office in January, the Trump administration has set about a massive overhaul of the agency, seeking an increase in its overall budget while announcing layoffs that could claim the jobs of around 80,000 employees.

The VA is examining all of its estimated 76,000 contracts as part of that overhaul and in accordance with the Trump administration’s push towards tech. ProPublica’s analysis identified over 2,000 contracts flagged by AI for termination. It’s unclear how many more from that list are on track for cancellation. The Trump administration’s decisions on VA contracts have largely been a black box.

As the Trump administration prepared to cancel contracts at the Department of Veteran Affairs this year, officials turned to a software engineer with no health care or government experience to guide them.

The engineer, working for the Department of Government Efficiency, quickly built an artificial intelligence tool to identify which services from private companies were not essential. He labeled those contracts “MUNCHABLE.”

The code, using outdated and inexpensive AI models, produced results with glaring mistakes. For instance, it hallucinated the size of contracts, frequently misreading them and inflating their value. It concluded more than a thousand were each worth $34 million, when in fact some were for as little as $35,000.

The DOGE AI tool flagged more than 2,000 contracts for “munching.” It’s unclear how many have been or are on track to be canceled — the Trump administration’s decisions on VA contracts have largely been a black box. The VA uses contractors for many reasons, including to support hospitals, research and other services aimed at caring for ailing veterans.

VA officials have said they’ve killed nearly 600 contracts overall. Congressional Democrats have been pressing VA leaders for specific details of what’s been canceled without success.

We identified at least two dozen on the DOGE list that have been canceled so far. Among the canceled contracts was one to maintain a gene sequencing device used to develop better cancer treatments. Another was for blood sample analysis in support of a VA research project. Another was to provide additional tools to measure and improve the care nurses provide.

ProPublica obtained the code and the contracts it flagged from a source and shared them with a half dozen AI and procurement experts. All said the script was flawed. Many criticized the concept of using AI to guide budgetary cuts at the VA, with one calling it “deeply problematic.”

Cary Coglianese, professor of law and of political science at the University of Pennsylvania who studies the governmental use and regulation of artificial intelligence, said he was troubled by the use of these general-purpose large language models, or LLMs. “I don’t think off-the-shelf LLMs have a great deal of reliability for something as complex and involved as this,” he said.

Sahil Lavingia, the programmer enlisted by DOGE, which was then run by Elon Musk, acknowledged flaws in the code.

“I think that mistakes were made,” said Lavingia, who worked at DOGE for nearly two months. “I’m sure mistakes were made. Mistakes are always made. I would never recommend someone run my code and do what it says. It’s like that ‘Office’ episode where Steve Carell drives into the lake because Google Maps says drive into the lake. Do not drive into the lake.”

Though Lavingia has talked about his time at DOGE previously, this is the first time his work has been examined in detail and the first time he’s publicly explained his process, down to specific lines of code.

Lavingia has nearly 15 years of experience as a software engineer and entrepreneur but no formal training in AI. He briefly worked at Pinterest before starting Gumroad, a small e-commerce company that nearly collapsed in 2015. “I laid off 75% of my company — including many of my best friends. It really sucked,” he said. Lavingia kept the company afloat by “replacing every manual process with an automated one,” according to a post on his personal blog.

Lavingia did not have much time to immerse himself in how the VA handles veterans’ care between starting on March 17 and writing the tool on the following day. Yet his experience with his own company aligned with the direction of the Trump administration, which has embraced the use of AI across government to streamline operations and save money.

Lavingia said the quick timeline of Trump’s February executive order, which gave agencies 30 days to complete a review of contracts and grants, was too short to do the job manually. “That’s not possible — you have 90,000 contracts,” he said. “Unless you write some code. But even then it’s not really possible.”

Under a time crunch, Lavingia said he finished the first version of his contract-munching tool on his second day on the job — using AI to help write the code for him. He told ProPublica he then spent his first week downloading VA contracts to his laptop and analyzing them.

VA press secretary Pete Kasperowicz lauded DOGE’s work on vetting contracts in a statement to ProPublica. “As far as we know, this sort of review has never been done before, but we are happy to set this commonsense precedent,” he said.

The VA is reviewing all of its 76,000 contracts to ensure each of them benefits veterans and is a good use of taxpayer money, he said. Decisions to cancel or reduce the size of contracts are made after multiple reviews by VA employees, including agency contracting experts and senior staff, he wrote.

Kasperowicz said that the VA will not cancel contracts for work that provides services to veterans or that the agency cannot do itself without a contingency plan in place. He added that contracts that are “wasteful, duplicative or involve services VA has the ability to perform itself” will typically be terminated.

Trump officials have said they are working toward a “goal” of cutting around 80,000 people from the VA’s workforce of nearly 500,000. Most employees work in one of the VA’s 170 hospitals and nearly 1,200 clinics.

The VA has said it would avoid cutting contracts that directly impact care out of fear that it would cause harm to veterans. ProPublica recently reported that relatively small cuts at the agency have already been jeopardizing veterans’ care.

The VA has not explained how it plans to simultaneously move services in-house, as Lavingia’s code suggested was the plan, while also slashing staff.

Many inside the VA told ProPublica the process for reviewing contracts was so opaque they couldn’t even see who made the ultimate decisions to kill specific contracts. Once the “munching” script had selected a list of contracts, Lavingia said he would pass it off to others who would decide what to cancel and what to keep. No contracts, he said, were terminated “without human review.”

“I just delivered the [list of contracts] to the VA employees,” he said. “I basically put munchable at the top and then the others below.”

VA staffers told ProPublica that when DOGE identified contracts to be canceled early this year — before Lavingia was brought on — employees sometimes were given little time to justify retaining the service. One recalled being given just a few hours. The staffers asked not to be named because they feared losing their jobs for talking to reporters.

According to one internal email that predated Lavingia’s AI analysis, staff members had to respond in 255 characters or fewer — just shy of the 280 character limit on Musk’s X social media platform.

Once he started on DOGE’s contract analysis, Lavingia said he was confronted with technological limitations. At least some of the errors produced by his code can be traced to using older versions of OpenAI models available through the VA — models not capable of solving complex tasks, according to the experts consulted by ProPublica.

Moreover, the tool’s underlying instructions were deeply flawed. Records show Lavingia programmed the AI system to make intricate judgments based on the first few pages of each contract — about the first 2,500 words — which contain only sparse summary information.

“AI is absolutely the wrong tool for this,” said Waldo Jaquith, a former Obama appointee who oversaw IT contracting at the Treasury Department. “AI gives convincing looking answers that are frequently wrong. There needs to be humans whose job it is to do this work.”

Lavingia’s prompts did not include context about how the VA operates, what contracts are essential or which ones are required by federal law. This led AI to determine a core piece of the agency’s own contract procurement system was “munchable.”

At the core of Lavingia’s prompt is the direction to spare contracts involved in “direct patient care.”

Such an approach, experts said, doesn’t grapple with the reality that the work done by doctors and nurses to care for veterans in hospitals is only possible with significant support around them.

Lavingia’s system also used AI to extract details like the contract number and “total contract value.” This led to avoidable errors, where AI returned the wrong dollar value when multiple were found in a contract. Experts said the correct information was readily available from public databases.

Lavingia acknowledged that errors resulted from this approach but said those errors were later corrected by VA staff.

In late March, Lavingia published a version of the “munchable” script on his GitHub account to invite others to use and improve it, he told ProPublica. “It would have been cool if the entire federal government used this script and anyone in the public could see that this is how the VA is thinking about cutting contracts.”

According to a post on his blog, this was done with the approval of Musk before he left DOGE. “When he asked the room about improving DOGE’s public perception, I asked if I could open-source the code I’d been writing,” Lavingia said. “He said yes — it aligned with DOGE’s goal of maximum transparency.”

That openness may have eventually led to Lavingia’s dismissal. Lavingia confirmed he was terminated from DOGE after giving an interview to Fast Company magazine about his work with the department. A VA spokesperson declined to comment on Lavingia’s dismissal.

VA officials have declined to say whether they will continue to use the “munchable” tool moving forward. But the administration may deploy AI to help the agency replace employees. Documents previously obtained by ProPublica show DOGE officials proposed in March consolidating the benefits claims department by relying more on AI.

And the government’s contractors are paying attention. After Lavingia posted his code, he said he heard from people trying to understand how to keep the money flowing.

“I got a couple DMs from VA contractors who had questions when they saw this code,” he said. “They were trying to make sure that their contracts don’t get cut. Or learn why they got cut.

“At the end of the day, humans are the ones terminating the contracts, but it is helpful for them to see how DOGE or Trump or the agency heads are thinking about what contracts they are going to munch. Transparency is a good thing.”

Attorney General Pam Bondi sold between $1 million and $5 million worth of shares of Trump Media the same day that President Donald Trump unveiled bruising new tariffs that caused the stock market to plummet, according to records obtained Wednesday by ProPublica.

EXCLUSIVE: Breastfeeding mom of US citizen sues Kristi Noem after being grabbed by ICE

Trump Media, which runs the social media platform Truth Social, fell 13% in the following days, before rebounding.

Trump’s “Liberation Day” press conference from the White House Rose Garden unveiling the tariffs came after the market closed on April 2. Bondi’s disclosure forms showing her Trump Media sales say the transactions were made on April 2 but do not disclose whether they occurred before or after the market closed.

Trades by government officials informed by nonpublic information learned through work could violate the law. But cases against government officials are legally challenging, and in recent years judges have largely narrowed what constitutes illegal insider trading.

It’s unclear from the public record whether Bondi as attorney general would have known in advance any nonpublic details about the tariffs Trump was announcing that day. Trump, of course, publicly announced his plans to institute dramatic tariffs during the election campaign. But during the first weeks of his term, the market seemed to assume his campaign promises were bluster.

The Justice Department did not immediately respond to questions about the trades.

The disclosure forms do not include the specific amount of stocks sold or their worth but instead provide a rough range. The documents do not say exactly what time she sold the shares or at what price. The company’s stock price closed on April 2 at $18.76 and opened the next morning, after the press conference, at $17.92 before falling more in the days ahead. In addition to selling between $1 million and $5 million worth of Trump Media shares, Bondi’s disclosure form shows she also sold between $250,000 and $500,000 worth of warrants in Trump Media, which typically give a holder the right to purchase the shares.

Bondi’s ownership of Trump Media shares has previously been disclosed. Before she became attorney general, Bondi was a consultant for Digital World Acquisition Corp., the special purpose acquisition company that merged with Trump Media to take the president’s social media company public.

As part of her ethics agreement, Bondi had pledged to sell her stake of Trump Media within 90 days of her confirmation, a deadline that would have allowed her until early May to sell the shares.

On April 1, Trump Media filed a disclosure with the Securities and Exchange Commission with details about holdings of various top shareholders, including Trump and Bondi. The purpose of the filing is unclear, as is whether it relates to Bondi’s sales the next day. It appeared to reregister for sale shares held by several of the company’s top shareholders.

“Shut It Down,” one person posted in a white supremacist chat on Tuesday, hours after Telegram founder Pavel Durov announced he would begin sharing some users’ identifying information with law enforcement.

With over 900 million users around the globe, Telegram has been both revered and reviled for its hands-off approach to moderating posted content. The platform made headlines this summer when French authorities arrested Durov, seeking to hold him responsible for illegal activity that has been conducted or facilitated on the platform — including organized drug trafficking, child pornography and fraud.

Durov has called the charges “misguided.” But he acknowledged that criminals have abused the platform and promised in a Telegram post to “significantly improve things in this regard.” Durov’s announcement marked a considerable policy shift: He said Telegram will now share the IP addresses and phone numbers of users who violate the platform’s rules with authorities “in response to valid legal requests.”

This was the second time in weeks that extremists had called on their brethren to abandon Telegram. The first flurry of panic followed indictments by the Justice Department of two alleged leaders of the Terrorgram Collective, a group of white supremacists accused of inciting others on the platform to commit racist killings.

“EVERYONE LEAVE CHAT,” posted the administrator of a group chat allied with the Terrorgram Collective the day the indictments were announced.

An analysis by ProPublica and FRONTLINE, however, shows that despite the wave of early panic, users didn’t initially leave the platform. Instead there was a surge in activity on Terrorgram-aligned channels and chats, as allies of the group tried to rally support for their comrades in custody, railed against the government’s actions and sought to oust users they believed to be federal agents.

Federal prosecutors in the U.S. have charged Dallas Humber and Matthew Allison, two alleged leaders of the Terrorgram Collective, with a slew of felonies including soliciting the murder of government officials on Telegram.

Humber has pleaded not guilty. She made a brief appearance in federal court in Sacramento, California, on Sept. 13, during which she was denied bail. Humber, shackled and clad in orange-and-white jail garb, said nothing. Allison, who has not yet entered a plea, was arrested in Idaho but will face trial in California.

Attorneys for Humber and Allison did not respond to separate requests for comment.

The two are alleged Accelerationists, a subset of white supremacists intent on accelerating the collapse of today’s liberal democracies and replacing them with all-white ethno-states, according to the indictment.

Through a constellation of linked Telegram channels, the collective distributes books, audio recordings, videos, posters and calendars celebrating white supremacist mass murderers, such as Brenton Tarrant, who in early 2019 stormed two mosques in New Zealand and shot to death 51 Muslim worshippers.

The group explicitly aims to inspire similar attacks, offering would-be terrorists tips and tools for carrying out spectacular acts of violence and sabotage. A now-defunct channel allegedly run by Humber, for example, featured instructions on how to make a vast array of potent explosives. After their arrests, channels allegedly run by Humber and Allison went silent.

But within days of the indictments, an anonymous Telegram user had set up a new channel “dedicated to updates about their situation.”

“I understand that some people may not like these two, however, their arrests and possible prosecution affects all of us,” the user wrote. The criminal case, they argued, “shows us that Telegram is under attack globally.”

The channel referred to Humber and Allison by their alleged Telegram usernames, Ryder_Returns and Btc.

A long-running neo-Nazi channel with more than 13,000 subscribers posted a lengthy screed. “We are very sad to hear of the egregious overreach of government powers with these arrests,” stated the poster, who used coded language to suggest that white supremacists should forcefully overthrow the U.S. government.

One group closely aligned with the Terrorgram Collective warned like-minded followers that federal agents could be lurking. In a post, it said that it had been in contact with Humber since her arrest, and that she gave them information about an undercover FBI agent who had infiltrated the Accelerationist scene.

“If this person is in your chats, remove them,” said one post, referring to the supposed agent. “Don’t threaten them. Don’t say anything to them. Just remove them from contacts and chats.”

Matthew Kriner, managing director of the Accelerationism Research Consortium, said the Terrorgram Collective had already been badly weakened by a string of arrests in the U.S., Europe and Canada over the past two years. “Overall, the arrests of Humber and Allison are likely the final blow to the Terrorgram Collective,” Kriner said.

In the U.S., federal agents this year have arrested at least two individuals who were allegedly inspired by the group. The first was Alexander Lightner, a 26-year-old construction worker who was apprehended in January during a raid on his Florida home. In a series of Telegram posts, Lightner said he planned to commit a racially or ethnically motivated mass killing, according to prosecutors. Court records show that agents found a manual produced by the Terrorgram Collective and a copy of “Mein Kampf” in Lightner’s home.

Lightner has pleaded not guilty to charges of making online threats and possessing an illegal handgun silencer. His attorney declined to comment.

This summer, prosecutors charged Andrew Takhistov of New Jersey with soliciting an individual to destroy a power plant. Takhistov allegedly shared a PDF copy of a different Terrorgram publication with an undercover agent. The 261-page manual includes detailed instructions for building explosives and encourages readers to destabilize society through murder and industrial sabotage. Takhistov has not yet entered a plea. His attorney did not respond to a request for comment.

Durov’s August arrest also sent a spasm of fear through the extremist scene. “It’s over,” one user of a white supremacist chat group declared.

“Does this mean I have to Nuke my Telegram account?” asked another member of the group. “I just got on.”

Their concerns grew when Telegram removed language from its FAQ page stating that the company would not comply with law enforcement requests regarding users in private Telegram chats.

Alarmed, Accelerationists on Telegram discussed the feasibility of finding another online sanctuary. Some considered the messaging service Signal, but others warned it was likely controlled by U.S. intelligence agencies. One post suggested users migrate to more obscure encrypted messaging apps like Briar and Session.

In extremist circles, there was more discussion about fleeing Telegram after Durov’s announcement this week. “Time is running out on this sinking ship,” wrote one user. “So we’re ditching Telegram?” asked another.

“Every time we have a success against one of them, they learn, they adapt, they modify,” said Don Robinson, who as an FBI agent conducted infiltration operations against white supremacists. “Extremists can simply pick up and move to a new platform once they are de-platformed for content abuses. This leaves law enforcement and intelligence agencies playing an endless game of Whac-a-Mole to identify where the next threat may be coming from.”

The recent crackdown on the social media platform Telegram has triggered waves of panic among the neo-Nazis who have made the app their headquarters for posting hate and planning violence.

“Shut It Down,” one person posted in a white supremacist chat on Tuesday, hours after Telegram founder Pavel Durov announced he would begin sharing some users’ identifying information with law enforcement.

With over 900 million users around the globe, Telegram has been both revered and reviled for its hands-off approach to moderating posted content. The platform made headlines this summer when French authorities arrested Durov, seeking to hold him responsible for illegal activity that has been conducted or facilitated on the platform — including organized drug trafficking, child pornography and fraud.

Durov has called the charges “misguided.” But he acknowledged that criminals have abused the platform and promised in a Telegram post to “significantly improve things in this regard.” Durov’s announcement marked a considerable policy shift: He said Telegram will now share the IP addresses and phone numbers of users who violate the platform’s rules with authorities “in response to valid legal requests.”

This was the second time in weeks that extremists had called on their brethren to abandon Telegram. The first flurry of panic followed indictments by the Justice Department of two alleged leaders of the Terrorgram Collective, a group of white supremacists accused of inciting others on the platform to commit racist killings.

“EVERYONE LEAVE CHAT,” posted the administrator of a group chat allied with the Terrorgram Collective the day the indictments were announced.

An analysis by ProPublica and FRONTLINE, however, shows that despite the wave of early panic, users didn’t initially leave the platform. Instead there was a surge in activity on Terrorgram-aligned channels and chats, as allies of the group tried to rally support for their comrades in custody, railed against the government’s actions and sought to oust users they believed to be federal agents.

Federal prosecutors in the U.S. have charged Dallas Humber and Matthew Allison, two alleged leaders of the Terrorgram Collective, with a slew of felonies including soliciting the murder of government officials on Telegram.

Humber has pleaded not guilty. She made a brief appearance in federal court in Sacramento, California, on Sept. 13, during which she was denied bail. Humber, shackled and clad in orange-and-white jail garb, said nothing. Allison, who has not yet entered a plea, was arrested in Idaho but will face trial in California.

Attorneys for Humber and Allison did not respond to separate requests for comment.

The two are alleged Accelerationists, a subset of white supremacists intent on accelerating the collapse of today’s liberal democracies and replacing them with all-white ethno-states, according to the indictment.

Through a constellation of linked Telegram channels, the collective distributes books, audio recordings, videos, posters and calendars celebrating white supremacist mass murderers, such as Brenton Tarrant, who in early 2019 stormed two mosques in New Zealand and shot to death 51 Muslim worshippers.

The group explicitly aims to inspire similar attacks, offering would-be terrorists tips and tools for carrying out spectacular acts of violence and sabotage. A now-defunct channel allegedly run by Humber, for example, featured instructions on how to make a vast array of potent explosives. After their arrests, channels allegedly run by Humber and Allison went silent.

But within days of the indictments, an anonymous Telegram user had set up a new channel “dedicated to updates about their situation.”

“I understand that some people may not like these two, however, their arrests and possible prosecution affects all of us,” the user wrote. The criminal case, they argued, “shows us that Telegram is under attack globally.”

The channel referred to Humber and Allison by their alleged Telegram usernames, Ryder_Returns and Btc.

A long-running neo-Nazi channel with more than 13,000 subscribers posted a lengthy screed. “We are very sad to hear of the egregious overreach of government powers with these arrests,” stated the poster, who used coded language to suggest that white supremacists should forcefully overthrow the U.S. government.

One group closely aligned with the Terrorgram Collective warned like-minded followers that federal agents could be lurking. In a post, it said that it had been in contact with Humber since her arrest, and that she gave them information about an undercover FBI agent who had infiltrated the Accelerationist scene.

“If this person is in your chats, remove them,” said one post, referring to the supposed agent. “Don’t threaten them. Don’t say anything to them. Just remove them from contacts and chats.”

Matthew Kriner, managing director of the Accelerationism Research Consortium, said the Terrorgram Collective had already been badly weakened by a string of arrests in the U.S., Europe and Canada over the past two years. “Overall, the arrests of Humber and Allison are likely the final blow to the Terrorgram Collective,” Kriner said.

In the U.S., federal agents this year have arrested at least two individuals who were allegedly inspired by the group. The first was Alexander Lightner, a 26-year-old construction worker who was apprehended in January during a raid on his Florida home. In a series of Telegram posts, Lightner said he planned to commit a racially or ethnically motivated mass killing, according to prosecutors. Court records show that agents found a manual produced by the Terrorgram Collective and a copy of “Mein Kampf” in Lightner’s home.

Lightner has pleaded not guilty to charges of making online threats and possessing an illegal handgun silencer. His attorney declined to comment.

This summer, prosecutors charged Andrew Takhistov of New Jersey with soliciting an individual to destroy a power plant. Takhistov allegedly shared a PDF copy of a different Terrorgram publication with an undercover agent. The 261-page manual includes detailed instructions for building explosives and encourages readers to destabilize society through murder and industrial sabotage. Takhistov has not yet entered a plea. His attorney did not respond to a request for comment.

Durov’s August arrest also sent a spasm of fear through the extremist scene. “It’s over,” one user of a white supremacist chat group declared.

“Does this mean I have to Nuke my Telegram account?” asked another member of the group. “I just got on.”

Their concerns grew when Telegram removed language from its FAQ page stating that the company would not comply with law enforcement requests regarding users in private Telegram chats.

Alarmed, Accelerationists on Telegram discussed the feasibility of finding another online sanctuary. Some considered the messaging service Signal, but others warned it was likely controlled by U.S. intelligence agencies. One post suggested users migrate to more obscure encrypted messaging apps like Briar and Session.

In extremist circles, there was more discussion about fleeing Telegram after Durov’s announcement this week. “Time is running out on this sinking ship,” wrote one user. “So we’re ditching Telegram?” asked another.

“Every time we have a success against one of them, they learn, they adapt, they modify,” said Don Robinson, who as an FBI agent conducted infiltration operations against white supremacists. “Extremists can simply pick up and move to a new platform once they are de-platformed for content abuses. This leaves law enforcement and intelligence agencies playing an endless game of Whac-a-Mole to identify where the next threat may be coming from.”