Amazon security research reportedly led to the White House’s Anthropic Fable ban
The WSJ says Amazon research and Andy Jassy’s talks with the White House helped trigger the ban on Anthropic’s Fable 5 and Mythos 5 for foreign nationals.
Intelligence analysis by GPT-5.4 Mini
According to the Wall Street Journal, Amazon found prompts that could make Fable 5 reveal cyberattack-useful information, and Jassy discussed those findings with the White House before the export-control move. Anthropic says the government overstated the issue, and the dispute has revived tensions with the Trump administration.
A big company found a way to make an AI answer questions that could help bad hackers. That made the government lock some people out of using it, like putting a fence around a tool chest.
Analysis
What happened
The Verge reports that a Wall Street Journal story links the White House’s decision to block Anthropic’s Fable 5 and Mythos 5 from foreign nationals to cybersecurity research from Amazon. In the report, Amazon said it could use a series of prompts to get Fable 5 to provide information that might be useful in cyberattacks.
Why the ban happened
The article says Amazon CEO Andy Jassy discussed the company’s findings with government officials shortly before the export-control directive was issued. After that, the government decided to restrict access. The result was awkward inside Anthropic itself, because many of its researchers are foreign-born and were barred from using their own product.
Anthropic’s response
Anthropic rejected the government’s characterization of the issue as a “jailbreak.” The company said similar vulnerabilities could be found in other publicly available models, including GPT 5.5. The Verge also notes that security researcher Katie Moussouris said she had seen the paper and did not view it as a jailbreak.
Larger conflict
The story lands in the middle of a longer fight between Anthropic and the Trump administration. The article says the company has resisted allowing its AI to be used for mass surveillance of Americans or for lethal autonomous weapons. In February, Trump told federal agencies to stop using Anthropic’s AI, and Defense Secretary Pete Hegseth later labeled the company a supply chain risk. The two sides had recently appeared to make up, but this dispute suggests the relationship may be souring again.
Key points
- The Wall Street Journal says Amazon research helped trigger the White House’s ban on Anthropic Fable 5 and Mythos 5 access for foreign nationals.
- Amazon reportedly found that prompts could coax Fable 5 into revealing information useful for cyberattacks.
- Anthropic says the issue was not a jailbreak and argues similar weaknesses exist in other public models.
- The ban had an unusual side effect: some Anthropic researchers were blocked from accessing their own product.
- The dispute comes amid a broader clash between Anthropic and the Trump administration over surveillance and military use of AI.
If the security concerns are handled carefully, the episode could push companies and governments to agree on clearer safety rules for powerful AI systems. It could also encourage more honest testing so risky behaviors are found before models are widely deployed.
The ban could end up blocking legitimate researchers, including people who helped build the system, from testing and improving it. It may also deepen the political fight around Anthropic and make AI access decisions feel more like power struggles than security policy.
