On June 13, the two most advanced models from Anthropic were put on pause by the U.S. government.

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

After reviewing all the details regarding this matter, we have roughly outlined the timeline of the past 24 hours.

On Thursday, June 11, just two days after the public release of Fable 5, Amazon CEO Andy Jassy raised concerns to the White House. He worried that the safety guards of Fable 5 could be bypassed. Amazon researchers reportedly used a series of prompts to have Fable 5 provide 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. 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 there was a three-way call.

Anthropic CEO Dario Amodei entered the call, facing about a half dozen senior officials. Besides Bessent and Cairncross, there was Commerce Secretary Howard Lutnick. Other participants included Deputy Secretary of Commerce Jeffrey Kessler, White House Staff Secretary Will Scharf, White House Deputy Chief of Staff Richard Walters, and presidential policy advisor Walker Barrett.

Amodei tried to explain the situation as a misunderstanding. He believed that what Amazon discovered was a specific method to bypass the system, not a general prompt jailbreak that could broadly dismantle safety guards. Anthropic later publicly stated that testers had not found a method to widely bypass the model's safety system.

However, the White House was not convinced.

Amazon's CEO's findings were sent to the National Security Agency for assessment, and White House officials believed they had enough evidence. The government requested that Anthropic voluntarily take down the models and work with the government to fix the vulnerabilities. Amodei wanted more time and information but did not commit to taking down the models. Bessent directly stated on the call that he made a "wrong decision."

Subsequently, export controls were imposed.

Anthropic offered another narrative. They claimed that the White House only gave 90 minutes to take the models offline and did not specify actual threat details. The White House stated that export controls were a last resort after hours of unsuccessful attempts to get Anthropic to comply.

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

By the end of 2024, Amazon made an additional $4 billion investment in Anthropic, raising the total investment amount to $8 billion. Anthropic simultaneously made AWS its primary training partner, and future model training and deployment would utilize AWS chips. Claude has also been one of the most important models on Amazon Bedrock.

Microsoft's alliance with OpenAI is well-known, and Amazon's bet on Anthropic was originally a way to sidestep that.

Microsoft has OpenAI. Google has Gemini and also invests in Anthropic. Amazon does not have a sufficiently strong in-house cutting-edge model and 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 had explored investing up to $50 billion in OpenAI. At the time, OpenAI was seeking up to $100 billion in new funding, and the potential deal could include OpenAI procuring Amazon AI chips. Axios also mentioned that OpenAI's annualized revenue would exceed $20 billion by 2025, but spending commitments would reach $14 trillion.

Amazon needs cutting-edge model companies to consume AWS computing power, validate in-house chips, fill data centers, and also wants to place the strongest models on its enterprise cloud shelves. This is no longer just a financial investment.

So it both invests in Anthropic and draws closer to OpenAI. It is both a financier for the model companies and a supplier to them. It needs to help them sell models while also explaining to the government how dangerous these models are.

In terms of results, Amazon found itself on the opposite side of Anthropic this time. In Anthropic's view, a partner that provides funding, cloud, chips, and distribution channels submitted a security signal to the government sufficient to trigger a ban. Of course, Amazon's own stance is, "The White House asked me, and I just responded to their questions."

Over the past two years, AI companies have liked to package themselves as national assets. The stronger the capability, the higher the valuation, the smoother the financing, the more imaginative the government procurement. Anthropic, in particular, excels at this narrative. It uses more cautious security language to distinguish itself from OpenAI while using "frontier risks" jargon to prove to regulators that it should be taken seriously.

Now, the U.S. government is indeed treating models as national security assets.

The confusion among White House officials also comes from here. Politico reported that the White House heard Amodei compare the dangers of Anthropic's technology to a nuclear bomb. When he refused to take down the model due to a known security vulnerability, government officials did not see it as a technical disagreement but rather a matter of attitude.

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

Anthropic and the U.S. government have had past conflicts.

This time, Anthropic stated that the government directive did not specify particular national security concerns and criticized the action for lacking transparency, clarity, and a legally defined process based on technical facts. Anthropic believes the issue is more akin to a narrowly defined bypass method, insufficient to support such a broad ban.

But from the government's perspective, model safety is no longer simply an internal process where the company writes white papers, conducts red team tests, and publishes system cards. Who can access the model, who can train the model, whether foreign employees can see model weights—all of these will fall into the language of export control.

When Anthropic announced in April that Mythos would only be open to limited tech and cybersecurity companies, it had already held multiple rounds of meetings with the White House. Before the launch of Fable 5, it also went through reviews from the U.S. government and the U.K. AI safety research institute. Anthropic believes that the government did not raise objections before the model was released.

This made the conflict more awkward.

Before the model release, it was a safety collaboration. After the model release, it became national security.

OpenAI observed this incident from the sidelines.

With Anthropic forced to take down its strongest model, OpenAI's relative position became more comfortable. The more entangled Anthropic is with regulatory issues, the easier it becomes for OpenAI to be seen as the "cooperative" option. If Amazon truly continues to draw closer to OpenAI, it will also have an added layer of hedging.

Of course, there is no public evidence showing that Amazon is trying to help OpenAI take down Anthropic.

The sharper fact is that as cutting-edge models enter a trillion-dollar capital expenditure cycle, partnerships are no longer clean. Cloud providers invest in model companies, model companies purchase cloud computing power, governments inquire about security risks from cloud providers, and competitors conduct red team tests in the same regulatory space.

Financiers, suppliers, distributors, and reviewers are increasingly played by the same set of companies.

This is more important than some prompt jailbreak.

On 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 still rests on the AWS console. OpenAI's financing table has not yet cleared. The U.S. government has already taken a front-row seat at the model launch event.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。