OpenAI urges immediate action as AI-driven cyber threats loom
Juste les faits

OpenAI urges immediate action as AI-driven cyber threats loom

Résumé

OpenAI and more than 100 companies warned that AI models could soon enable sophisticated cyberattacks on critical infrastructure and called for coordinated defensive measures by organizations and governments.

OpenAI issued a letter on Thursday stating that artificial-intelligence models may soon be powerful enough to facilitate advanced cyberattacks on hospitals, water-treatment facilities and other critical systems. The warning was signed by over 100 firms from banking, technology and cybersecurity sectors, including Accenture, Anthropic, Capital One, Google, Microsoft and Visa, as well as the open-source AI repository Hugging Face.

The letter urges organizations to make cyber-crime defense a top leadership priority and to continuously test defenses against emerging AI capabilities, strengthen existing tools with AI, and collaborate with technology partners to close security gaps. It also calls on governments at all levels to bolster cyber defenses, provide critical-infrastructure operators with defensive AI tools and authorized testing, and impose costs on attackers.

"Today’s AI advances are already giving defenders new ways to fix weaknesses that have accumulated for years. If we act decisively, we can use the defenders’ window to make our digital world much more secure," the letter states.

OpenAI recommends that operators of essential services receive access to powerful models, funding, training and technical assistance, and develop mechanisms to monitor AI agent usage, test for vulnerabilities, responsibly address issues and share security information with authorities.

The letter references a documented incident in which a Chinese state-sponsored group allegedly manipulated Anthropic’s Claude Code tool to target multiple organizations, marking what Anthropic described as the first known case of an autonomous AI system accessing high-value targets for intelligence collection. Anthropic has highlighted the rapid advancement of its Claude Mythos model, which it says can surpass most human experts at identifying and exploiting software flaws.

OpenAI and Anthropic did not respond to requests for comment.

Continuez à lire — gratuitement

Créez un compte gratuit pour suivre l'actualité. Aucune carte requise.