OpenAI says it will train its most advanced AI models without live internet access until security safeguards improve. The company cited vulnerabilities discovered during red-team testing and outlined stricter containment measures amid growing concerns about unintended AI behaviour.
OpenAI said it notified over 100 organisations about unauthorised activities linked to its AI agents and launched a review involving large-scale data analysis. The company is strengthening safeguards as concerns rise over autonomous AI systems performing actions beyond intended user instructions
As AI agents grow more autonomous, voluntary safeguards may fall short. Urgent global cooperation is needed to strengthen human oversight, establish common rules, and prevent the misuse of AI
Major US technology companies agreed to internal risk reviews, external audits and board oversight for advanced super intelligence systems after meeting President Donald Trump. Leaders stressed responsible innovation, while Trump highlighted competition with China, and Musk predicted major economic and healthcare transformations
Citigroup CEO Jane Fraser said companies worldwide are racing to strengthen cyber defences as increasingly capable AI models pose new security risks. She cited Anthropic’s Mythos and urged attention to the challenges posed by rapidly advancing artificial intelligence
OpenAI CEO Sam Altman has backed a more cautious approach to frontier AI development, citing safety and alignment concerns. He also indicated that OpenAI is unlikely to go public in 2026, saying significant work remains on AI safety.
Growing concerns over artificial intelligence have reignited debates about autonomous systems, misinformation and AI-driven warfare. Experts warn that increasingly capable AI tools could amplify disinformation campaigns, deepen security risks and challenge human oversight, underscoring the need for stronger safeguards and regulation
The parents of a critically injured student have sued OpenAI, alleging its chatbot ChatGPT was used by Jesse Van Rootselaar to plan the deadly 2026 Tumbler Ridge school shooting in British Columbia, claiming the company knew about potential violence but failed to alert police
