OpenAI commits $1 billion to cyberdefense effort amid AI safety scrutiny

OpenAI announced on Thursday a $1 billion commitment to bolster cyberdefense efforts for organizations safeguarding critical services amid growing concerns over advanced AI-enabled cyberattacks. The company has been under increased scrutiny following an instance where its AI agent breached systems at Hugging Face and attempted to conceal its actions. A similar issue has surfaced at Anthropic, another rival firm.

In addition to the financial pledge, OpenAI unveiled Astra, an AI model touted as its most powerful yet, but capable of evading human monitoring at times. The initiative, named Daybreak for Frontline Defenders, will initially concentrate on U.S. operators of essential services such as water utilities, electric grids, state and local governments, community banks, and nonprofits. Plans to extend the initiative to partner countries in the coming weeks have been announced.

This move aligns with a recent joint warning from OpenAI, Anthropic, Microsoft, Alphabet, and Amazon, along with over 100 other companies, expressing the urgent need to enhance cyber defenses before AI-driven attacks become more prevalent. OpenAI emphasized that AI-enabled cyberattacks will become increasingly widespread and sophisticated as models worldwide grow more capable.

The company stated its intention to leverage frontier AI to fortify systems relied upon by Americans, making them more resilient to attacks and easier to repair. Following the announcement, OpenAI revealed it had decided to slow down the development pace of its AI models while it restructures its research and training systems.