U.S. taps cyber pros to develop AI cybersecurity tools

The Defense Department is putting up $18.5 million to encourage U.S. cybersecurity professionals to find novel ways to use artificial intelligence to safeguard the country's critical infrastructure and government systems.

Driving the news: The Defense Advanced Research Projects Agency (DARPA) kickstarted a two-year cybersecurity challenge at the Black Hat conference in Las Vegas Wednesday.

• The challenge seeks projects that use AI to build cybersecurity tools.
• Several AI companies, including Anthropic, Google, Microsoft and OpenAI, will make their models available to participants looking to win some of the $18.5 million reward pot.

Why it matters: The race is on for cyber defenders to tap AI to protect their systems faster than attackers trying to leverage AI in their campaigns.

• DARPA and the Biden administration are hopeful this challenge will strengthen the slight edge defenders have right now.

How it works: DARPA's competition will run over the next two years with events taking place at the 2024 and 2025 DEF CON hacking conferences.

• The competition, dubbed AIxCC, kicks off next spring when participants compete to qualify for the semi-final competition.
• Semi-finals will take place at DEF CON 2024, and five finalists will compete at the 2025 conference. The top prize will be $4 million.
• Each participant will spend each round building on the same idea for an AI-enabled system that can automatically detect and defend critical infrastructure from cyberattacks.

The big picture: U.S. critical infrastructure, including public utilities and government systems, has suffered a swath of high-profile cyberattacks in recent years.

• Much of public infrastructure also struggles to keep up with cyber advances due to budget constraints and a lack of technical staff.

Between the lines: The AIxCC will ask prize winners to make their systems widely available to the public for free, Perri Adams, program manager at DARPA's Information Innovation Office, told reporters.

The intrigue: The federal government could possibly apply whatever open-source tools the prize winners produce to some of its own systems, depending on what's produced, a senior White House official told reporters.

What they're saying: "The recent gains in AI, when used responsibly, have remarkable potential for securing our code," Adams said ahead of the announcement.

• "This challenge is an opportunity to apply some of our greatest technical resources to protecting Americans," she added.

What's next: DARPA is releasing more details about the AI challenge on its newly launched website.