Skip to main content

Google, OpenAI team up with the White House for cybersecurity challenge

Google, OpenAI team up with the White House for cybersecurity challenge


A handful of companies are also opening their systems up to public review.

Share this story

An illustration of a woman typing on a keyboard, her face replaced with lines of code.
Image: The Verge

The Biden administration is partnering with four top artificial intelligence companies to sponsor a new cybersecurity challenge aimed at protecting the US’s critical infrastructure.

The “AI Cyber Challenge,” announced at the Black Hat USA conference in Las Vegas Wednesday, is meant to pair experts with AI models produced by Anthropic, Google, Microsoft, and OpenAI to develop systems to identify and fix software vulnerabilities. The challenge, hosted by DARPA, includes $20 million in prizes for the best systems that could be used in protecting the government infrastructure running everything from transportation to electrical grids.

“They don’t have the tools capable of security at this scale.”

“In an increasingly interconnected world, software undergirds everything from public utilities to our financial systems,” Perri Adams, DARPA program manager, said in a call with reporters Tuesday. “Cyber defenders are tasked with protecting a daunting maze of technology, and today, they don’t have the tools capable of security at this scale.”

Teams are expected to start competing in a qualifying event next spring, and the winners will go on to compete in a semifinal event at next year’s Def Con. The top five teams will go on to compete at the finals, which will be held at Def Con 2025. Prize winners will then be asked to open source their systems so they “can be used by everyone from volunteer, open-source developers to commercial industry,” Adams said Tuesday.

Adams and other challenge coordinators are expected to hold a keynote, virtual press conference, and panel later this week to provide additional details.

In May, Google, Microsoft, Nvidia, OpenAI, and Meta agreed to open their language models up for public evaluation “by thousands of community partners and AI experts” at this year’s Def Con starting later this week. 

The new challenge comes after a series of cybersecurity announcements from the White House this week. On Tuesday, the Biden administration held a summit on K-12 ransomware attacks, announcing a series of actions to mitigate the threat. The Federal Communications Commission is expected to launch a pilot program to provide $200 million in subsidies over the next three years for K-12 schools and libraries to bolster their cyber defenses. Amazon Web Services also pledged $20 million to fund a grant program for schools and state departments of education. 

During Tuesday’s press call, deputy national security advisor for cyber and emerging technology, Anne Neuberger, said the challenge would bolster US defenses against foreign actors.