Anthropic Relaunches Claude Mythos 5, Fable 5 Anticipated

Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
Anthropic Relaunches Claude Mythos 5, Fable 5 Expected
Suspended by Anthropic since June 12 to comply with a directive from the U.S. government, the powerful AI Claude Mythos 5 is back at certain organizations. The news was announced by its developer in a statement published on X. However, this redeployment is limited to a group of organizations, and importantly, it only concerns U.S. organizations.
“Today, the government informed us that Mythos 5, our most advanced cybersecurity model, could be redeployed to a set of U.S. organizations responsible for operating and protecting critical infrastructure,” reads the announcement. Additionally, rumors suggest that Claude Fable 5 (the consumer version of Claude Mythos 5) could also return this week. For reference, Claude Mythos 5 is a highly effective AI in cybersecurity, capable of identifying hundreds of security vulnerabilities in software, and is reserved for selected organizations that use it for defensive purposes.
The developer of Claude also announces that it continues to work with the U.S. government to expand access to Claude Mythos 5 and to make Fable 5 available to the general public. As a reminder, Claude Fable 5 is the consumer version of Claude Mythos 5. Fable 5 has the same performance levels as Mythos 5, but its capabilities in cybersecurity, biology, and chemistry have been restricted for safety reasons.
Fable 5 had to be suspended because the U.S. government was convinced that there is already a way to unrestrict Claude Fable 5 for malicious purposes. Nevertheless, Anthropic asserts that this technique does not actually allow for the unrestricting of Claude Fable 5. “We reviewed a demonstration of this specific technique used to identify a small number of already known minor vulnerabilities. These vulnerabilities all appear relatively simple, and we found that other publicly accessible models can also detect them without the need for a workaround,” Anthropic argued in a blog post.
Anthropic indicates that it has the green light from Washington to redeploy Claude Mythos 5 at a few U.S. organizations. Rumors suggest that Claude Fable 5, the consumer version, could also return this week. Anthropic continues to discuss with the U.S. government to expand the redeployment of Claude Mythos 5 and restore public access to Claude Fable 5.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.