Trustworthy AI for National Security - Kathleen Fisher - PSW #805

AI/ML is providing significant benefits in a wide range of application domains but also provides adversaries with a new attack surface. Learn about DARPA's efforts to help evaluate AI/ML and work towards a trust model that will allow us to use these valuable tools safely. Segment Resources: Identifying and Mitigating the Security Risks of Generative AI paper (co-authored by Kathleen): https://arxiv.org/abs/2308.14840 DARPA’s AI Forward, which will include AI Exploration opportunities and resource material: https://www.darpa.mil/work-with-us/ai-forward I2O webpage, important to include because this hosts links to many of the programs Dr. Fisher will discuss: https://www.darpa.mil/i2o In the Security News: If an exploit falls in the forest do I still need to patch?, Reflections on trusting trust: the source code revealed, prompt injection in your resume, iPhones be updating, a deep dive into vulnerable kernel drivers and wiping SPI flash, cheap to exploit software, to ransom or steal?, oh OAuth, Florida man, door bell shenanigans, don’t pay the ransom, the White House and AI, and quantum teleportation via measurement-induced entanglement. All that and more on this episode of Paul’s Security Weekly! Visit https://www.securityweekly.com/psw for all the latest episodes! Follow us on Twitter: https://www.twitter.com/securityweekly Like us on Facebook: https://www.facebook.com/secweekly Show Notes: https://securityweekly.com/psw-805