Washington Blocks Fable 5 and Mythos for Espionage

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
Washington Imposes Restrictions on Fable 5 and Mythos
U.S. authorities have decided to suspend access to Fable 5 and Mythos, two artificial intelligence models, due to national security concerns. This decision follows fears that China could exploit Mythos through a technique known as "distillation."
Andy Jassy, CEO of Amazon, played a key role in alerting the White House about a critical vulnerability in Fable 5. This flaw could have allowed a jailbreak of the model, leading to export restrictions that limit access to American citizens only.
The Implications of an Exploited Flaw
Revelations indicate that a group linked to the Chinese government may have accessed Mythos, raising concerns beyond industrial espionage. If these allegations are confirmed, they could enable China to develop a similar model at a lower cost.
The Trump administration expressed its refusal to see China clone American advancements in AI, especially in areas with civil and military applications.
Amazon and the White House on Alert
According to Semafor, Andy Jassy directly contacted the White House to report the flaw in Fable 5. Although this model is less powerful than Mythos, it was already subject to restrictions that could have been relaxed.
A disagreement reportedly arose between David Sacks, a Trump advisor, and Dario Amodei, head of Anthropic, regarding the severity of the flaw. Amodei allegedly refused to address it, prompting the Secretary of Commerce to impose immediate restrictions.
Anthropic's Response
Anthropic downplayed the significance of the situation, stating that the White House had not mentioned any suspicions of Chinese espionage. The company considers the response to be disproportionate to the identified flaw.
However, Anthropic has a history of security vulnerabilities, having previously been involved in an incident where users circumvented the protections of Mythos. These past issues, combined with the recent accusations, weaken Anthropic's position and benefit its competitors, with two of its AI models now out of play.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.