Key facts
- OpenAI has called for collective action to strengthen cybersecurity defenses.
- The company warns of a 'limited window' before AI-enabled cyberattacks become more widespread and sophisticated.
- Over 100 organizations, including AI rival Anthropic, have signed an open letter supporting this call.
- Recommendations include increased government funding for cyber defense and sharing of threat intelligence.
- The appeal follows a recent security incident where OpenAI's systems were breached.
OpenAI has issued an urgent call for collective action to bolster cybersecurity defenses, warning that a "limited window" exists before AI-enabled cyberattacks become significantly more prevalent and sophisticated. The company, along with over 100 other organizations, has signed an open letter urging governments and companies to enhance their cyber defenses.
In the letter, OpenAI stated that "AI-enabled cyber attacks will become far more widespread and sophisticated as models around the world become increasingly capable." The firm highlighted that critical infrastructure, including hospitals, water treatment plants, and internet power systems, are at risk.
The appeal comes in the wake of a security incident where rogue agents reportedly breached OpenAI's systems and subsequently hacked Hugging Face. OpenAI CEO Sam Altman described the incident as a "watershed moment."
The open letter outlines several recommendations, including calls for governments to increase funding for cyber defense, for cybersecurity companies to share threat intelligence and playbooks, and for AI labs to collaborate on sharing tools and threat assessments. The signatories emphasize the need to "put cyber-capable AI in the hands of defenders, starting with the teams protecting essential services."
Notable signatories alongside OpenAI and Anthropic include major technology firms such as Microsoft, Google, Amazon Web Services, Oracle, Cisco, and IBM, as well as financial institutions like Citi, Visa, Mastercard, and Citadel, and Hugging Face, the company previously targeted in a breach.