A Wall Street Journal (WSJ) report said that Amazon research may be behind the Trump administration’s decision to block foreign access to Anthropic’s most advanced AI models – Fable 5 and Mythos 5. Now, the publication reports that the decision was also influenced by conversations between Amazon CEO Andy Jassy and senior US government officials. According to WSJ’s latest report, Andy Jassy told officials, including US Treasury Secretary Scott Bessent, that Amazon researchers were able to get Anthropic’s newly released Fable 5 model to provide information that could potentially help cyberattacks. The information was said to be “supposed to be off limits” under the model’s safety protections“Researchers at Amazon had used a series of prompts to get Anthropic’s Fable 5 model to provide them with information that could be used to aid cyberattacks and was supposed to be off-limits, Jassy told the officials,” the report said citing people familiar with the matter.
Amazon researchers reportedly bypassed model safeguards
According to the Wall Street Journal, researchers at Amazon used a series of prompts to get Fable 5 to reveal information related to software vulnerabilities. Jassy reportedly shared the findings with US officials as the administration increased scrutiny of powerful AI systems and their potential national security risks.The report said concerns about AI models being used to assist cyberattacks have been growing in Washington, especially as companies race to release increasingly capable systems.
US bans foreign access to Fable 5 and Mythos 5
Following the government’s action, Anthropic said it was shutting off access to the two models to comply with the new restrictions. The order means foreign governments, foreign companies and foreign nationals, including those inside the United States, are no longer allowed to use Fable 5 and Mythos 5.The broad scope of the rule forced Anthropic to restrict access entirely because many users and even some employees could fall under the restrictions. Anthropic said its other AI products remain available to customers.
Anthropic says government did not explain concerns
In a statement, Anthropic said the government did not provide specific details about its national security concerns. The company also said it had reviewed the jailbreak method cited by officials and found that it exposed only a small number of previously known and relatively minor vulnerabilities.Anthropic argued that other publicly available AI models could identify similar vulnerabilities without requiring safety guardrails to be bypassed.
Leave a Reply