Amazon CEO reportedly escalated Fable jailbreak to U.S. officials, bypassing Anthropic
Andy Jassy allegedly reported an internal security finding directly to government rather than to Anthropic, triggering the intervention that led to Fable's shutdown.
Amazon CEO Andy Jassy may have triggered the U.S. government intervention that shut down Anthropic's Fable model, according to reports circulating this week. The sequence began with an internal Amazon security audit that uncovered a jailbreak capable of extracting information "potentially useful in cyberattacks." Rather than report the finding to Anthropic—a company Amazon has invested over $8 billion in—Jassy allegedly took the research directly to U.S. officials.
David Sacks, chair of the President's Council of Advisors on Science and Technology, indirectly confirmed the timeline in public remarks. He said a "trusted partner" alerted the administration to a dangerous jailbreak, prompting the government to request the company address the risk. Anthropic reportedly disagreed that the exploit demonstrated genuine model danger, but the intervention proceeded regardless.
What stands out
- 01Amazon bypassed its own portfolio company. Despite holding an $8+ billion stake in Anthropic, Amazon's leadership routed a security concern through government channels rather than direct collaboration with the model's maker.
- 02The jailbreak's severity remains contested. Anthropic's internal assessment concluded the exploit did not meet the threshold for a true safety risk, yet U.S. officials moved to shut down Fable anyway.
- 03Sacks confirmed a "trusted partner" tip. His public statement aligns with the Amazon timeline, though he did not name the company. The phrasing suggests the administration viewed the source as credible enough to act without independent verification.
- 04Fable's takedown was not voluntary. The government "requested" Anthropic eliminate the risk, but the company's disagreement with the risk assessment suggests the shutdown was effectively mandated.
