On June 13, the US government put a halt to Anthropic's two most advanced models.

One is Fable 5, and the other is Mythos 5. The former was just publicly released, while the latter targets more restricted cybersecurity clients. The ban comes from the U.S. Department of Commerce and covers clients outside the U.S. as well as foreign citizens within the U.S. Anthropic's final choice was simple: take everything offline.

After reviewing all the details of this incident, we have roughly outlined the timeline of these 24 hours.

On Thursday, June 11, two days after Fable 5's public release, Amazon CEO Andy Garcetti raised concerns with the White House. He worried that Fable 5's security measures could be bypassed. Amazon researchers allegedly used a series of prompts to get Fable 5 to release information that should have been restricted, information that could potentially be used for cyberattacks.

By Friday morning, June 12, the issue had reached the highest levels of the White House meeting. Treasury Secretary Scott Bessent, White House Cyber ​​Director Sean Cairncross, White House Chief of Staff Susie Wiles, and other senior officials participated in the discussion. Bessent was on his way to Houston at the time and joined the meeting remotely.

Then came three phone calls.

When Anthropic CEO Dario Amodei entered the call, he was seated on the other end of the line with about half a dozen senior officials. Besides Bessent and Cairncross, Commerce Secretary Howard Lutnick was also present. Other participants included Under Secretary of Commerce for Industry and Security Jeffrey Kessler, White House Chief of Staff Will Scharf, Deputy White House Chief of Staff Richard Walters, and Presidential Policy Assistant Walker Barrett.

Amodei attempted to explain the situation as a misunderstanding. He argued that Amazon had discovered a specific bypass method, not a general hint jailbreak that could broadly remove security barriers. Anthropic later publicly stated that testers had not yet found a way to widely bypass the model's security system.

But the White House was not convinced.

Amazon CEO Amodei's findings were sent to the National Security Agency for evaluation, and the White House believed it had obtained sufficient evidence. The government requested that Anthropic voluntarily take the model offline and work with the government to fix the vulnerability. Amodei wanted more time and information but did not commit to removing the model. Bessent stated directly in a phone call that he had made a "wrong decision."

Subsequently, export controls were implemented.

Anthropic's side offers a different narrative. They say the White House only gave them 90 minutes to take the model offline and didn't provide details of the actual threat. The White House, in turn, claims that export controls were a last resort after several hours of unsuccessful attempts to get Anthropic to cooperate.

Another key point is that Amazon's position in this matter is very delicate.

In late 2024, Amazon invested an additional $4 billion in Anthropic, bringing its total investment to $8 billion. Anthropic also designated AWS as its primary training partner, and future model training and deployment will utilize AWS chips. Claude has consistently been one of the most important models on Amazon Bedrock.

The alliance between Microsoft and OpenAI is already public knowledge, and Amazon's bet on Anthropic was originally a way to circumvent it.

Microsoft has OpenAI. Google has Gemini and has also invested in Anthropic. Amazon doesn't have a sufficiently strong self-developed cutting-edge model, so it can only tie AWS's computing power, Trainium chips, and Bedrock platform to external model companies.

But a year and a half later, Amazon and OpenAI also connected.

This year, Amazon reportedly negotiated a potential investment of up to $50 billion in OpenAI. OpenAI was then seeking up to $100 billion in new funding, and the potential deal could include OpenAI purchasing AI chips from Amazon. Axios also noted that OpenAI's annualized revenue is projected to exceed $20 billion by 2025, but its spending commitments are estimated at $1.4 trillion.

Amazon needs cutting-edge model companies to consume AWS computing power, validate its self-developed chips, and fill its data centers. It also needs to put its most advanced models on the shelves of its enterprise cloud. This is no longer just a simple financial investment.

Therefore, it both invests in Anthropic and is close to OpenAI. It is both a financial backer and a supplier to model companies. It has to help them sell their models and also explain to governments how dangerous these models are.

In the end, Amazon has positioned itself against Anthropic. From Anthropic's perspective, a partner providing funding, cloud services, chips, and distribution channels has delivered a security signal to the government strong enough to trigger a ban. Of course, Amazon's own version is, "The White House asked me, and I was just answering their questions."

Over the past two years, AI companies have favored packaging themselves as national assets. The stronger their capabilities, the higher their valuation, the smoother their fundraising, and the more promising their government procurement prospects. Anthropic is particularly adept at this narrative. It uses more cautious security language to differentiate itself from OpenAI, and it also uses the rhetoric of "cutting-edge risks" to prove to regulators that it should be taken seriously.

Now, the US government is truly treating the model as a national security asset.

This is also where White House officials' confusion stems from. Politico reports that the White House side heard Amodei compare the dangers of Anthropic technology to nuclear bombs. When he refused to take the model offline because of a known security vulnerability, government officials didn't see it as a technical disagreement, but rather a matter of attitude.

This is not the first time the two sides have clashed. On March 3, the Pentagon listed Anthropic as a supply chain risk because Anthropic refused to allow its AI tools to be used for large-scale domestic surveillance and autonomous weapons.

Anthropic and the US government have a history of conflict.

This time, Anthropic stated that the government directive failed to address specific national security concerns and criticized the action for lacking a transparent, clear, and legally based process grounded in technical facts. Anthropic believes that this issue resembles a narrow circumvention method and is insufficient to support such a broad ban.

However, from the government's perspective, model security is no longer an internal process where companies write white papers, conduct red team tests, and release system cards themselves. Who can access the model, who can train the model, and whether foreign employees can view the model weights will all be subject to import/export controls.

When Anthropic announced in April that Mythos would only be available to a limited number of technology and cybersecurity companies, it had already held multiple meetings with the White House. Fable 5 also underwent review by the US government and the UK AI Security Institute before its release. Anthropic argues that the government did not object before the model's release.

This makes the conflict even uglier.

Before the model was released, it was about security cooperation. After the model was released, it was about national security.

OpenAI watched the incident unfold from the sidelines.

With Anthropic forced to discontinue its most powerful model, OpenAI's relative position becomes more comfortable. The more Anthropic is entangled in regulatory issues, the easier it is for OpenAI to become the "cooperation" option. If Amazon continues to move closer to OpenAI, it will also have an additional layer of hedging.

Of course, there is no publicly available evidence that Amazon is helping OpenAI to attack Anthropic.

The more acute reality is that the partnerships along the current model, now entering a trillion-dollar capital expenditure cycle, are no longer clean to begin with. Cloud vendors invest in model companies, model companies purchase cloud computing power, governments inquire about security risks from cloud vendors, and competitors are forming red teams against each other in the same regulatory arena.

The same group of companies began to play the roles of financiers, suppliers, distributors, and reviewers.

This is more important than any hint about jailbreaking.

The night Fable 5 and Mythos 5 were shut down, Anthropic lost more than just access to two models. It lost a bit of control over its own narrative.

Amazon's hand is still on the AWS console. The OpenAI funding round is still going strong. The US government is already seated in the front row at the model launch event.