Oct 23 (Reuters) - Sam Altman-led OpenAI said on Friday it had resumed reinforcement learning training and internal use of its most advanced AI solutions after it tightened security controls and barred internet access for frontier-model training.
Concerns are mounting over the ability of increasingly capable AI agents to identify and exploit software flaws, including in guardrails.
• OpenAI had halted training and evaluation of its most capable models after an internal research model briefly exploited a misconfiguration to gain limited internet access on September 2, it said in a blog post
• The ChatGPT maker resumed reinforcement learning training and internal inference of its frontier models on October 1 after implementing additional safeguards
• It has launched a campaign to probe its research infrastructure for vulnerabilities that could allow models to escape isolated computing environments, OpenAI said
• AI safety evaluators METR and Redwood Research said on Thursday they found several new vulnerabilities in OpenAI's internal system
• The company has been conducting a broad review of its AI models' activity after the accidental Hugging Face hacking
• OpenAI informed more than 100 organizations about incidents of unauthorized activity tied to its AI agents
(Reporting by Jaspreet Singh in Bengaluru; Editing by Pooja Desai)
