Anthropic Revises Its Policy on Claude Fable 5 in Response to Criticism

Anthropic Revises Its Strategy on Claude Fable 5

Anthropic, a key player in the field of artificial intelligence, has recently canceled a policy it had implemented for its Claude Fable 5 model. This decision follows negative reactions from the AI community. Initially, the company had introduced secret protections to limit the use of this advanced model, but it had to reassess its position after facing criticism.

Claude Fable 5, launched this week, is a public version of the Mythos model, designed with additional security measures to prevent misuse. At its launch, Anthropic had put in place precautions such as redirecting questions about cybersecurity, biology, and chemistry to less capable models, like Opus 4.8, to ensure that the advanced model would not be used to plan cyberattacks or build biological weapons.

Transparency and Adjustments

Anthropic also indicated that for those attempting to use Fable 5 for AI development, the company would degrade the model's performance without explaining the change to the user. This measure was perceived by some members of the developer community as a discreet attempt to prevent the creation of competing AI systems.

In response to the criticism, Anthropic decided to make these protections more transparent. From now on, Fable 5 will inform users if their requests are denied or redirected. A company spokesperson stated that all requests flagged on the API will provide a reason for their denial, marking a significant change in the company's communication policy.

Anthropic acknowledged that it had made "the wrong trade-off" and apologized for not finding the right balance initially. The company emphasized that its protective measures aim to address national security issues, particularly to prevent foreign adversaries from advancing in the development of advanced technologies. It clarified that a large majority of coding and machine learning work is not affected by these protections.

The Mythos Model Under Scrutiny

Meanwhile, Anthropic's Mythos model, announced in April, continues to be a source of concern for governments and security agencies. Considered one of the most powerful AI systems, Mythos is capable of performing advanced reasoning and cybersecurity tasks that exceed the capabilities of current public models. Researchers and Anthropic themselves have reported that Mythos could be used to accelerate cyberattacks, assist in research on biological or chemical weapons, and provide foreign actors with a new dangerous tool. Due to its potential, the model is distributed in a limited manner to approved government users, rather than to the general public.