Topic: preparedness framework
-
OpenAI restricts Astra over critical cyber capabilities
OpenAI's forthcoming model Astra may reach a "critical" cybersecurity capability threshold, triggering stricter internal safeguards and a more cautious deployment strategy under its Preparedness Framework. Preliminary evaluations show Astra could autonomously identify unknown vulnerabilities or o...
Read More » -
OpenAI delays next model over critical cyber risk
OpenAI's stress tests on its upcoming model Astra were strong enough that the company "cannot rule out" it possessing top-tier critical cyber capabilities, prompting a pause on select internal development and expanded security protocols. "Critical" is the highest risk tier in OpenAI's Preparednes...
Read More » -
OpenAI Seeks New Head of Preparedness to Lead Safety Efforts
OpenAI is recruiting a Head of Preparedness to lead proactive identification and mitigation of catastrophic AI risks, from cybersecurity threats to long-term dangers, as part of its formal Preparedness Framework. CEO Sam Altman highlighted specific concerns driving this focus, including AI models...
Read More » -
OpenAI halts major AI training run over cyber threat risks
OpenAI paused reinforcement learning training on its frontier models for two weeks due to cybersecurity concerns and a recent security incident, using the time to harden research environments, expand monitoring, and conduct red-team evaluations. The halt follows the OpenAI-Hugging Face incident a...
Read More » -
OpenAI warns of 'high' AI weaponization risk, unveils countermeasures
OpenAI warns of a high risk of AI weaponization for cyberattacks and is implementing countermeasures to safeguard its technology while aiding defenders. AI's rapid advancement, demonstrated by a sharp increase in performance on cybersecurity challenges, suggests it could soon develop sophisticate...
Read More » -
OpenAI Fortifies AI Defenses Against Rising Cyber Threats
AI capabilities in cybersecurity are advancing rapidly, with OpenAI's models showing a dramatic performance increase, which could enable more sophisticated cyber operations. Experts emphasize that strong foundational security practices remain the best defense, as AI amplifies existing threats and...
Read More » -
OpenAI hit the brakes. What happens next?
OpenAI announced a deliberate slowdown in AI development, including a two-week halt on reinforcement learning training for deployment-bound models, to strengthen security protocols,a notable voluntary pause amid intense competitive pressure from rivals like Anthropic and Chinese developers. The p...
Read More »