OpenAI Raises "Critical" Risk with Astra
OpenAI associates its Astra model with an unprecedented level of cybersecurity and does not rule out a transition to a "critical" risk zone. During evaluations, two unknown vulnerabilities are said to have been discovered. The company communicated about this upcoming model two days ago.
Two Unknown Vulnerabilities and a Considered "Critical" Risk
OpenAI states that Astra has identified two vulnerabilities that were not known during its evaluations. The company specifies that the model's capabilities are such that it can no longer exclude crossing its threshold of risk classified as "critical." This level is presented as unprecedented in the field of cybersecurity for OpenAI.
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.
Choisissez votre rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
What the "Critical" Threshold Entails at OpenAI
The risk level classified as "critical" applies when a system is capable of identifying and exploiting zero-day vulnerabilities. It also encompasses the ability to plan and execute a sophisticated cyberattack largely without human intervention. Execution autonomy is part of the criteria defining this threshold.
A Model Announced, Presented as the Most Advanced in Security
Two days ago, OpenAI published a post about Astra, presented as an upcoming model. The company unveiled the name GPT-6 Astra and describes it as its most advanced and powerful model in cybersecurity.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.