Supporting competitors for DARPA’s AI Cyber Challenge

One of the clearest benefits of artificial intelligence (AI) is the potential to enhance cybersecurity for organizations and internet users around the world. This is especially true as bad actors continue to target and exploit critical software and systems. Advances in AI will reshape these dynamics, and responsibly applying AI could help take the industry toward a more secure architecture for the digital world.

Experience also tells us that getting there requires strong collaborations between the public and private sector, as well as opportunities for innovation like DARPA’s AI Cyber Challenge (AIxCC) Semifinal Competition event at DEF CON 32 in Las Vegas from August 8-11. The two-year challenge brings together AI and cybersecurity researchers to build new AI tools to help secure major open-source projects.

This is the inaugural competition of the challenge which DARPA announced at last year’s Black Hat. The Semifinal Competition comes on the heels of the anniversary of the White House’s voluntary AI commitments – where industry and government came together to agree to advance responsible practices in the development and use of AI. Today, we’re sharing how Google is supporting AIxCC competitors along the way:

• Google Cloud resources: To set up competitors for success, Google is providing$5,000 in Google Cloud credits for each qualifying AIxCC team, up to $1 million in total. Participants can use the credits for eligible Google Cloud services, including Gemini. This means competitors will have access to Google Cloud AI and machine learning models, which they can use and build upon for various tasks during the challenge. We also encourage participants to take advantage of our incentive programs, like the Google for Startups Program, which covers Google Cloud costs up to $350,000 for AI startups.
• Cybersecurity experts: Google has been pioneering using AI for security for years, whether for advanced threat detection or protecting Gmail users from phishing. We’ve seen the potential of AI for security and look forward to sharing what we’ve learned with AIxCC by making our security experts available throughout the challenge. Google security experts will be able to provide advice on projects and infrastructure, and help develop the baseline for evaluating participants. To kick things off, we published a guide on how AI can enhance Google's open source OSS-Fuzz technology, which specifically shares recommended approaches to AIxCC.
• AIxCC experience at DEF CON: Next week, we'll be at the AIxCC Semifinal Experience at DEF CON with an AI-focused hacker demo and hands-on demos to showcase Google's AI technology. We will also have Chromebooks at the Google booth to demo our AI technology including Google Security Operations and SecLM. Google security experts will be on hand to share knowledge and participate in technical discussions with attendees.

AI Village at DEF CON 32

In addition to our work with the AIxCC, Google will continue to support the AI Village at DEF CON, which educates participants on security and privacy issues around AI. This year, Google donated $10,000 to the Village and will be providing Pixelbooks for workshop attendees to get hands-on experience at the intersection of AI and cybersecurity. The AI Village is returning after last year’s successful AI Village red teaming event, which Google took part in and where security professionals worked on AI security challenges.

We look forward to seeing the new innovations that stem from the AIxCC Semifinal Competition, and sharing how they can be applied to secure the software we all rely on. In the meantime, you can learn more about our AI security work via the Secure AI Framework and the recently announced Coalition for Secure AI.