Right this moment at Black Hat USA convention, the  Protection Superior Analysis Tasks Company (DARPA) and Open Supply Safety Basis (OpenSSF) introduced the AI Cyber Problem (AIxCC), a two-year competitors to construct synthetic intelligence that may routinely detect and repair safety vulnerabilities in software program.

Cybercriminals appear to have a endless provide of safety vulnerabilities at their disposal, giving them alternatives to worm their means into the delicate laptop programs and databases maintained by the world’s largest enterprises and governments.

DARPA and OpenSSF intention to place a dent in that parade of vulnerabilities with the AIxCC, which guarantees $18.5 million in prize cash in addition to entry to AI know-how from OpenAI, Anthropic, Google, and Microsoft.

“Open supply software program is an important and core a part of our nation’s crucial infrastructure,” mentioned OpenSSF Common Supervisor Omkhar Arasaratnam mentioned in a launch launch. “Discovering new and modern methods to make sure our open supply software program provide chain is safe by building is in everybody’s greatest curiosity.”

The competitors will function an open observe and a funded observe. People who find themselves excited about taking part within the AIxCC open observe can register with DARPA beginning in November. DARPA will even choose seven small companies to compete within the funded observe; companies excited about receiving funding to compete can register beginning August 17.

There can be two phases to the AIxCC, together with semifinal competitors and the ultimate competitors. The 2 occasions can be held at DEF CON in Las Vegas in 2024 and 2025, respectively. As much as 20 groups can be chosen to take part within the semifinal, whereas as much as 5 can be chosen for the finals. For extra info, go to www.aicyberchallenge.com.

Associated Objects:

Feds Increase Cyber Spending as Safety Threats to Information Proliferate

Safety Issues Inflicting Pullback in Open Supply Information Science, Anaconda Warns

Hacking AI: Exposing Vulnerabilities in Machine Studying