The United States government ordered Anthropic to cut off foreign access to two of its most advanced AI models, Fable 5 and Mythos 5, over national security concerns. The Commerce Department issued the directive on June 1, 2026, in a letter signed by Commerce Secretary Howard Lutnick to Anthropic CEO Dario Amodei, requiring government approval for any export, re-export, or domestic transfer of the models to non-U.S. persons. Anthropic confirmed compliance on June 13, 2026, and disabled both models for all customers worldwide, though access to the company's other AI models remains unaffected. Amazon's cloud unit AWS confirmed that Anthropic asked it to revoke access for all users in all regions.

The Commerce Department cited the models' advanced cybersecurity capabilities and the risk that they could be weaponized by hostile actors as the basis for the controls. According to Anthropic, the government believes it has identified a way to bypass or break through the safety restrictions built into the Fable 5 model, specifically a safeguard meant to prevent the model from being used to identify software vulnerabilities. The action reportedly followed claims that another company had managed to break through the safety restrictions on the Mythos model. Anthropic said the government provided only "verbal evidence of a potential narrow, non-universal jailbreak" and disagreed that such a finding should be cause for recalling a commercial model deployed to hundreds of millions of people.

The restriction applies even to foreign nationals located inside the United States, meaning a non-U.S. individual in the U.S. can no longer use the models. Before the shutdown, access to the Mythos models was limited to a program called Project Glasswing, which provided the technology mainly to U.S. companies such as Microsoft, Google, and NVIDIA for use as digital immune systems to address critical cybersecurity vulnerabilities. Some Indian entities, including the Indian Cybercrime Coordination Centre and CERT-in, had also gained limited access through Project Glasswing earlier in June 2026, and that access is now threatened.

Anthropic expressed concern that there may have been a misunderstanding and said it is working to restore access as quickly as possible. The company warned that if this standard were applied across the industry, it would essentially halt all new model deployments for every major AI provider. The Pentagon's chief information officer, Kirsten Davies, said the Defense Department supported prioritizing national security, adding that some things are more important than revenue cycles and pre-IPO valuation. Anthropic had confidentially filed for a U.S. IPO last month.

The tech community reacted with surprise and criticism. Dean Ball with the Foundation for American Innovation called the move baffling and cartoonish, questioning why an administration that supports exporting advanced AI chips to China would simultaneously block foreign access to the best American models. Peter Girnus with Zero Day Initiative noted that Anthropic had spent months marketing Mythos as dangerously powerful and compared the situation to 1990s encryption export controls, which ultimately collapsed. Marc Andreessen of Andreessen Horowitz and Chris McGuire with the Council on Foreign Relations both expressed concern that broad export controls without warning are questionable, and McGuire argued that the current export control strategy is incoherent and self-defeating. Dan Shipper, CEO of Every, predicted the ban would be lifted within days. Jeremy Howard, cofounder of fast.ai, criticized both the government decision and Anthropic for not anticipating the consequences of marketing a model as too dangerous for non-Americans to use. Ryan Brewer of OpenAI warned that continued restrictions could eventually limit frontier AI access to only a few buildings in the Bay Area.

Cybersecurity researchers at Hudson Rock argued that the restrictions are already causing a surge in cybercriminal activity on the darknet and that expected identity verification measures will ultimately fail to keep cybercriminals out. They pointed to infostealer malware such as Lumma, Vidar, and RedLine, which harvest active session tokens and credentials from infected computers, including access to major AI platforms. Hudson Rock identified over 30,000 corporate OpenAI credentials captured through infostealer infections. Underground markets also exist for pre-verified AI platform accounts, and services offer deepfake generation and real-time voice manipulation to defeat biometric liveness checks. The researchers also warned that requiring AI companies to collect and store large volumes of sensitive personal documents creates high-value targets for hackers.

The order comes amid an already strained relationship between Anthropic and the U.S. government. Earlier this year, the company refused to allow the U.S. military to use its AI models for domestic surveillance and fully autonomous weapons systems. In response, the Pentagon labeled Anthropic a supply chain risk and moved to end its role as a military AI provider. That dispute had recently shown signs of easing before the new directive was issued. The action marks a significant escalation of U.S. efforts to restrict foreign adversaries' access to advanced artificial intelligence, establishing a precedent that the government will treat advanced AI models as strategic assets similar to high-technology semiconductors or military equipment.

Original Sources/Tags: infostealers.com, timesofindia.indiatimes.com, cryptobriefing.com, wired.com, h25.io, yahoo.com, indiatoday.in, thehindu.com, (mythos), (openai), (banks), (cryptocurrency), (cookies), (hackers), (regulations)

This article provides limited practical value to a normal person. It describes a cybersecurity problem involving AI export controls and criminal activity on the darknet, but it does not offer any steps, tools, or choices that a reader can act on. There are no instructions to follow, no resources mentioned, and no clear actions a person can take based on this information. The article simply relays what researchers at Hudson Rock have found about stolen credentials, underground markets, and methods criminals use to bypass identity checks. A person who reads this cannot apply it to their own life in any direct way, unless they happen to work in cybersecurity, AI policy, or law enforcement, in which case the only relevant information is the broad outline of the threats described.

The educational depth is moderate but uneven. The article introduces several important concepts, such as infostealer malware, session hijacking, synthetic identity fraud, and the use of deepfakes to defeat biometric checks. It explains how stolen session tokens and cookies can bypass multifactor authentication, which is a useful technical insight. It also draws a parallel between cryptocurrency exchange security and the expected identity verification measures for AI platforms, which helps the reader understand the pattern. However, the article does not explain how infostealer malware actually infects computers in the first place, what specific technical controls could reduce the risk of session hijacking, or how biometric liveness checks work at a technical level. The mention of 30,000 corporate OpenAI credentials is presented without context for how that number compares to the total number of credentials in circulation, so the reader cannot judge whether that figure represents a small fraction or a significant portion. The article teaches the reader that the threat exists and is serious, but it leaves major gaps for someone who wants to understand the full picture or evaluate the claims critically.

Personal relevance is narrow for most people. The story directly affects cybersecurity professionals, AI company employees, corporate IT administrators, and people who manage access controls for sensitive systems. For those individuals, the information is meaningful and could influence decisions about security policies or vendor choices. For everyone else, the relevance is indirect. The article touches on broader themes like identity theft, data breaches, and the security of personal information, which are important topics. But it does not explain how likely an average person is to be affected by stolen AI platform credentials, what to do if they are concerned about their own accounts being compromised, or how to evaluate whether a particular security measure is trustworthy. The relevance is limited to people who follow cybersecurity news closely or who have a personal stake in AI platform security.

The public service function is weak. The article does not offer warnings, safety guidance, or emergency information that helps the public act responsibly. It does not tell readers how to protect their own accounts from infostealer malware, what to do if they suspect their credentials have been stolen, or how to evaluate the security practices of the platforms they use. The article appears to exist mainly to report on a research finding and argue that new regulations will fail, rather than to help people make better decisions or stay safe. It informs but does not guide.

There is no practical advice in the article. No steps or tips are given that an ordinary reader can follow. The guidance is entirely absent. The article does not even suggest general actions a person might take when thinking about their own cybersecurity or evaluating the risks of using AI platforms.

The long term impact is minimal for most readers. The information does not help a person plan ahead, improve habits, or make stronger choices. It focuses on a specific regulatory development and its expected consequences, with no lasting benefit for the average reader. However, for people who work in cybersecurity or AI policy, the article highlights the importance of designing security measures that account for existing criminal infrastructure, and it underscores the risk that collecting sensitive identity documents could create new targets for attackers. This is a useful lesson, but the article does not develop it into practical guidance.

The emotional and psychological impact is concerning. The article creates a sense of inevitability and helplessness by repeatedly emphasizing that criminal infrastructure is mature, tested, and profitable, and that new regulations will only slow actors down temporarily. The tone suggests that the problem is unsolvable, which may leave readers feeling that there is no point in trying to improve security. While the article is not alarmist in its language, the cumulative effect of its arguments is to make the threat feel overwhelming and the proposed solutions feel futile. A reader may finish the article feeling informed about a serious problem but uncertain about what, if anything, can be done.

There is no clickbait or ad driven language. The article is straightforward and does not use exaggerated or dramatic claims. It does not sensationalize the event or rely on shock to maintain attention. The tone is factual and analytical, though the subject matter itself is inherently concerning.

The article misses several chances to teach or guide. It presents a major security finding but fails to provide context, examples, or ways for the reader to learn more. It could have explained how individuals can protect themselves from infostealer malware, what steps companies can take to reduce the risk of credential theft, or what patterns exist in how underground markets operate. A reader who wants to learn more could compare independent news sources on the same topic, look for analysis from established cybersecurity organizations or academic institutions, or consider general media literacy practices like checking whether multiple credible outlets report the same findings and looking for the original research rather than relying on a single summary.

To add real value, a reader can take several practical steps based on general reasoning and universal security principles. When using any online platform that requires login credentials, enable multifactor authentication whenever possible, as this adds a layer of protection even if a password is stolen. Be cautious about downloading software or clicking links from unknown sources, since infostealer malware often spreads through phishing emails, fake downloads, and compromised websites. Use a unique, strong password for each important account, and consider using a reputable password manager to keep track of them. If you receive a notification that your credentials have appeared in a data breach, change your passwords immediately and check for any unauthorized activity on your accounts. When evaluating whether a platform or service takes security seriously, look for clear information about their security practices, such as whether they offer multifactor authentication, how they store personal data, and whether they have a history of responding quickly to breaches. If you are concerned about your personal information being used for identity fraud, monitor your credit reports and financial accounts regularly for unexpected changes, and consider placing a fraud alert or credit freeze with major credit bureaus if you suspect your information has been compromised. These steps are realistic, widely applicable, and grounded in common sense. They help readers protect their own accounts and personal information, even though the original article offered none of this guidance.

The text says "cybercriminals have spent years perfecting methods to bypass bank-level identity checks." The word "perfecting" makes their skills sound advanced and polished, like a craft. This helps the story feel more dramatic by making the criminals seem very good at what they do. It pushes the reader to feel that the threat is serious and hard to stop.

The text says "dedicated criminal operations have routinely found ways around them." The word "dedicated" makes the criminals sound focused and hardworking, almost like they have a real job. This gives them a strange kind of respect in the story. It helps the message that these groups are hard to beat.

The text says "a threat actor in a restricted country can buy these stolen logs for a small fee." The phrase "a small fee" makes the cost sound tiny and easy. This hides how serious it is by making it sound like anyone can do it without much effort. It helps the idea that access to these tools is cheap and simple.

The text says "hijack a legitimate user's active session, completely bypassing any identity verification." The word "legitimate" makes the real user sound good and innocent. This pushes the reader to feel sorry for the real user and angry at the criminal. It helps the story by making the harm feel personal and unfair.

The text says "Hudson Rock's own intelligence has identified over 30,000 corporate OpenAI credentials." The phrase "own intelligence" makes Hudson Rock sound like a spy agency with special knowledge. This helps them seem more trustworthy and important. It pushes the reader to believe their numbers without question.

The text says "representing a massive pool of exploitable access." The word "massive" is a strong word that makes the problem sound very big. This pushes the reader to feel worried and to think the threat is larger than it might be. It helps the story feel urgent and serious.

The text says "a well-established underground market already exists for pre-verified AI platform accounts." The phrase "well-established" makes the market sound old and solid, like a real business. This hides how wrong it is by making it sound normal and organized. It helps the idea that this problem has been around for a long time.

The text says "treating access to restricted models as a readily tradable commodity." The word "commodity" makes the stolen accounts sound like regular goods, like food or clothes. This hides the crime by making it sound like a normal trade. It helps the story by making the market feel like a real economy.

The text says "criminals turn to advanced synthetic identity fraud." The word "advanced" makes the criminals' tools sound high-tech and smart. This pushes the reader to feel that the criminals are always one step ahead. It helps the message that new rules will not work.

The text says "deepfake generation and real-time voice manipulation specifically designed to defeat biometric liveness checks." The phrase "specifically designed to defeat" makes the tools sound like they were built for one clear purpose. This pushes the reader to feel that the criminals are very focused and skilled. It helps the story by making the threat feel targeted and hard to stop.

The text says "bad actors already have an inexpensive, complete toolkit to fabricate verified identities." The word "inexpensive" makes the tools sound cheap and easy to get. This hides how serious the problem is by making it sound simple. It helps the idea that many people could do this, not just experts.

The text says "there is also an ironic security risk in the push for stricter identity checks." The word "ironic" tells the reader how to feel about the situation before explaining it. This pushes the reader to see the new rules as foolish or self-defeating. It helps the message that stricter checks will backfire.

The text says "requiring AI companies to collect and store large volumes of sensitive personal documents creates high-value targets for hackers." The phrase "high-value targets" makes the databases sound like treasure chests. This pushes the reader to feel that the new rules will cause more harm than good. It helps the story by making the solution sound like a new problem.

The text says "the stolen credentials will flow directly back into the cybercriminal ecosystem." The word "ecosystem" makes the criminal world sound like a natural system, like a forest or a pond. This hides the harm by making it sound like a normal part of life. It helps the story by making the criminal world feel organized and stable.

The text says "the infrastructure to circumvent AI access restrictions already exists, has been tested, and is actively profitable." The phrase "actively profitable" makes the criminal work sound like a real business that earns money. This pushes the reader to feel that the problem is not going away because it pays well. It helps the message that new rules will not stop people who make money from this.

The text says "the new regulations may slow some actors down temporarily." The word "temporarily" makes the rules sound weak and short-lived. This pushes the reader to feel that the new rules are not worth much. It helps the story by making the regulations seem like a small speed bump, not a real solution.

The text says "the tools and markets to get around them are mature and readily available." The word "mature" makes the criminal tools sound fully grown and ready, like fruit that is ripe. This pushes the reader to feel that the problem is already as bad as it can get. It helps the message that there is no easy fix.

The text only talks about what cybercriminals can do and what might go wrong with new rules. It does not talk about any good that the export controls might do, like slowing down harmful AI use by bad groups. This leaves out one side of the story. It helps the message that the rules are not working by only showing the problems.

The text uses Hudson Rock as its main source but does not say if Hudson Rock has any reason to push this story. It does not say if they sell security tools or services that would benefit from more fear. This hides whether the source might have its own goals. It helps the story seem neutral by not questioning the source.

The text says "cybercriminals" many times but never says which countries or groups they belong to. This hides who is really behind the activity. It helps the story by keeping the criminals vague, so no one group or place looks worse than another.

The text says "restricted country" but does not name the country. This hides which place is being talked about. It helps the story by not blaming one nation, but it also hides important details from the reader.

The text says "frontier AI systems" to describe the models. The word "frontier" makes the AI sound exciting and cutting edge. This pushes the reader to feel that these models are very important and worth protecting. It helps the story by making the stakes feel high.

The text says "AI companies are expected to rely on" certain checks. The passive voice hides who expects this or who is pushing for these rules. It helps the story by not naming the people or groups behind the new rules. This keeps the focus on the criminals, not on the rule makers.

The text says "the very measures AI companies are expected to rely on." The word "very" adds emphasis and makes the measures sound important. This pushes the reader to feel that these checks are the main hope, which then makes their failure seem bigger. It helps the story by building up the measures just to knock them down.

The text says "bad actors already have an inexpensive, complete toolkit." The phrase "bad actors" is a soft way to say criminals or hackers. This hides how serious their actions are by using a calm, almost polite term. It helps the story by making the criminals sound less scary, even though the text says they do harmful things.

The text says "the overall picture is one where the infrastructure to circumvent AI access restrictions already exists." The phrase "the overall picture" tells the reader to see this as the full truth. This pushes the reader to accept the story as complete and final. It helps the message by making the reader feel that there is nothing more to know.

The text does not talk about any times when identity checks have worked to stop bad activity. It only talks about how criminals get around them. This leaves out proof that checks can help. It helps the story by making it seem like checks never work.

The text says "cybercriminals have spent years perfecting methods to bypass bank-level identity checks." This makes it sound like banks have tried and failed for a long time. But the text does not say if banks have had any success at all. This hides any good results and only shows the failures. It helps the message that new checks on AI will also fail.

The text says "despite years of strict identity verification and anti-money laundering procedures on crypto exchanges, dedicated criminal operations have routinely found ways around them." The word "despite" sets up a contrast that makes the criminal wins sound bigger. This pushes the reader to feel that the rules are useless. It helps the story by making past efforts sound like a waste of time.

The text says "there is also an ironic security risk in the push for stricter identity checks." This sets up a strawman by making it sound like people who want stricter checks have not thought about the risks. The text does not say who is pushing for these checks or if they have thought about this problem. It helps the story by making the other side look careless without proving they are.

The text says "if these databases are breached, the stolen credentials will flow directly back into the cybercriminal ecosystem." The word "if" makes this sound like a guess, but the text presents it as almost certain. This pushes the reader to feel that a breach is very likely, even though it has not happened yet. It helps the story by making the risk feel real and close.

The text says "the new regulations may slow some actors down temporarily." The word "may" makes this sound unsure, but the text treats it as fact that the rules will not last. This pushes the reader to feel that the rules are weak. It helps the message by making the regulations seem like they will fail.

The text says "the tools and markets to get around them are mature and readily available." This makes it sound like there is no way to stop the criminals because everything they need is already out there. This pushes the reader to feel hopeless about the problem. It helps the story by making the reader think that new rules are pointless.

The text does not talk about any good reasons for the export controls, like stopping harmful AI use by dangerous groups. It only talks about how criminals will get around them. This leaves out the other side of the debate. It helps the message by making the rules seem like a bad idea without showing why they were made.

The text says "U.S. government export controls that pulled the AI models Mythos and Fable offline are already causing a surge in cybercriminal activity on the darknet." This makes it sound like the controls caused the surge, but the text does not prove this. It pushes the reader to blame the rules for the problem. It helps the story by making the regulations look like they made things worse.

The text says "cybercriminals have spent years perfecting methods to bypass bank-level identity checks using synthetic identities and mule accounts." This makes it sound like the criminals are very skilled and always ahead. It pushes the reader to feel that they cannot be stopped. It helps the story by making the threat seem too big to handle.

The text says "a threat actor in a restricted country can buy these stolen logs for a small fee." The phrase "restricted country" hides which country it is. This keeps the reader from knowing who might be behind the activity. It helps the story by not pointing fingers at any one place.

The text says "Hudson Rock's own intelligence has identified over 30,000 corporate OpenAI credentials captured through infostealer infections." This uses a big number to make the problem sound very serious. But the text does not say how many total credentials exist, so the reader cannot tell if 30,000 is a lot or a little. It helps the story by making the number sound scary without giving full context.

The text says "the stolen credentials will flow directly back into the cybercriminal ecosystem, providing exactly the materials needed to fuel further identity fraud and access bypasses." The word "exactly" makes it sound like the breach will definitely help criminals. This pushes the reader to feel that the new rules will cause more crime. It helps the story by making the solution look like a new problem.

The text expresses a strong sense of warning and concern throughout, which serves as its primary emotional driver. This feeling of alarm appears right from the start, where the writer says that export controls are "already causing a surge in cybercriminal activity." The word "surge" makes the problem sound fast and out of control, like a wave that is growing bigger. This sets the tone for the entire piece, telling the reader that something bad is happening right now and getting worse. The purpose of this emotion is to make the reader pay attention and feel that the topic is urgent, that this is not a small or distant problem but one that is unfolding in the present moment.

A feeling of frustration or even defeat runs through the text, especially when the writer talks about how criminals keep finding ways around security measures. The phrase "dedicated criminal operations have routinely found ways around them" carries a tone of weariness, as if the writer is saying that no matter what people do, the criminals always win. The word "routinely" makes it sound like this happens over and over again, which adds to the sense that the problem cannot be fixed. This emotion serves to prepare the reader for the argument that new rules will not work, because the same thing has already failed in other areas like cryptocurrency exchanges. It guides the reader to feel that trying harder with the same tools is pointless.

The text also creates a feeling of fear around the idea that personal information could be stolen and misused. When the writer says that requiring AI companies to collect "large volumes of sensitive personal documents" creates "high-value targets for hackers," the emotion here is one of dread. The phrase "high-value targets" makes the databases sound like treasure that criminals desperately want, which makes the reader worry about what happens if those databases are breached. The word "ironic" at the start of that section adds a layer of dark surprise, as if the solution has accidentally created a new problem. This emotion is meant to make the reader question whether stricter identity checks are really a good idea, because the very act of collecting more personal data could put that data at risk.

A sense of admiration, though not positive in this context, is directed at the criminals themselves. Words like "perfecting," "dedicated," and "advanced" make the criminals sound skilled and organized, almost like professionals at the top of their field. The phrase "a well-established underground market already exists" gives the criminal world a feeling of being solid and permanent, like a real business that has been around for a long time. This is not admiration in the sense of approval, but rather a way of making the threat feel bigger and more serious. By making the criminals sound capable and organized, the writer pushes the reader to feel that these are not small-time troublemakers but a powerful force that is hard to stop. This emotion serves to build the argument that new regulations will not be enough because the opposition is too strong.

The text expresses a feeling of inevitability, a sense that the outcome is already decided. Phrases like "the infrastructure to circumvent AI access restrictions already exists, has been tested, and is actively profitable" carry a tone of finality. The word "already" appears multiple times and reinforces the idea that everything the criminals need is in place right now, that the problem is not coming in the future but is here today. The phrase "the tools and markets to get around them are mature and readily available" adds to this feeling by making the criminal tools sound fully developed and easy to get. This emotion of inevitability is perhaps the most powerful one in the text, because it pushes the reader to feel that the fight is already over before it has even started. It guides the reader toward the conclusion that the new regulations are not just likely to fail but are essentially pointless.

A subtle feeling of distrust toward the proposed solutions also runs through the text. The writer does not directly say that the people behind the regulations are wrong, but the repeated emphasis on failure and irony suggests that the solutions are poorly thought out. The phrase "the very measures AI companies are expected to rely on" builds up the importance of these measures only to knock them down, which creates a feeling of disappointment. The passive voice in "AI companies are expected to rely on" hides who is doing the expecting, which subtly distances the writer from the people pushing these rules. This emotion of distrust is meant to make the reader skeptical of the regulations and more sympathetic to the writer's argument that they will not work.

The writer uses several tools to increase the emotional impact of the text. Repetition is one of the most noticeable. The word "already" appears again and again, each time reinforcing the idea that the problem is here and now. The phrase "the tools and markets" is echoed in different forms throughout the text, creating a rhythm that makes the argument feel solid and well-supported. Comparisons are another key tool. The writer compares the expected identity checks for AI platforms to those used by banks and cryptocurrency exchanges, which gives the reader a familiar reference point. This comparison makes the argument easier to understand and more emotionally resonant, because the reader can see the pattern of failure repeating in a new context. The use of specific numbers, like "over 30,000 corporate OpenAI credentials," adds a feeling of precision and seriousness, making the threat feel real and measurable rather than abstract.

The writer also uses vivid language to make the criminal methods sound dramatic and threatening. Words like "hijack," "breached," and "exploitable" carry strong emotional weight, making the reader feel that these are not just technical problems but serious attacks. The phrase "silently harvest" to describe how infostealer malware works creates a feeling of sneakiness and danger, like a thief moving through a house without being seen. The description of deepfake tools as "specifically designed to defeat biometric liveness checks" makes the criminals sound like engineers building weapons, which adds to the sense that they are always one step ahead. These word choices are not neutral. They are carefully selected to make the reader feel worried and to build the case that the threat is too big for the proposed solutions to handle.

Overall, the emotions in the text work together to guide the reader toward a specific reaction. The warning and fear make the topic feel urgent. The frustration and inevitability make the proposed solutions seem futile. The distrust toward the regulations makes the reader question whether the people in charge know what they are doing. And the vivid descriptions of criminal capability make the threat feel overwhelming. The writer is not just presenting facts. The writer is using emotion to persuade the reader that the new export controls and identity verification rules are not just ineffective but potentially counterproductive, and that the real story is one of a problem that is already out of control.