Anthropic, the $965 billion artificial intelligence company, has revised its approach to safeguarding its Fable 5 model after facing backlash from researchers over hidden restrictions, according to a statement released this week. The company announced it would make its safety protocols more transparent, including visibly downgrading certain requests related to advanced AI development.
Why Anthropic Revisited Its Safeguards
The controversy began when a 319-page safety document revealed Fable 5 would silently downgrade requests from researchers building their own AI systems. This practice, which the company described as a “tradeoff” to prevent misuse, drew criticism from figures like Jeremy Howard, cofounder of Fast.ai, who argued it “slows down recursive AI self-improvement.” Anthropic’s head of product management, Dianne Na Penn, had previously stated the company felt “more confident with our safety guardrails in place” when releasing Fable 5 in April. However, the lack of transparency led to widespread concern, prompting the company to adjust its policies.
“We made the wrong tradeoff and we apologize for not getting the balance right,” an Anthropic spokesperson said in a statement to Fortune. The revised safeguards, effective this week, will ensure flagged requests “visibly fall back to Opus 4.8” and return a reason for their rejection via the API.
How the Shift Impacts AI Research
Anthropic’s decision to disclose when safeguards are triggered addresses a key demand from the AI research community. The company emphasized that its restrictions “do not affect the vast majority of coding and ML work,” but explicitly prohibit using Fable 5 to create competing AI systems. This policy aligns with industry norms, though critics argue it stifles innovation.
The move also highlights the growing intersection of AI safety and national security. Anthropic cited concerns that foreign adversaries could use its technology to enhance their own AI capabilities, particularly in optimizing chips developed by “those adversaries.” The company noted that U.S. and allied dominance in “frontier chips and optimized software” is a critical factor in its decision-making.
What’s Next for Anthropic?
The revision comes amid broader scrutiny of AI ethics and regulation. Earlier this year, Anthropic faced a standoff with the U.S. Department of Defense, which labeled the company a “supply chain risk” due to concerns over surveillance and autonomous weapons. While the Pentagon later limited defense contractors from using Anthropic’s models, the dispute remains unresolved.
Anthropic’s pivot also coincides with its confidential IPO filing, underscoring the tension between corporate growth and ethical responsibility. The company’s initial decision to obscure its safeguards “touched a nerve in the AI research community,” as one researcher noted, but its latest adjustments aim to balance transparency with security.
As AI systems grow more powerful, the debate over who controls their development—and how—will remain central to the industry’s future.