Phillip Misner
@phillipmisner.bsky.social
Head of AI Incident Detection & Response @ MSFT, ecosystem & customer advocate, incident responder, PSIRT enthusiast, safety & security.
AI jailbreaks are a common concern where attackers can influence the outcome of generative AI models. This week we released more guidance on how developers can protect against these threats: news.microsoft.com/source/featu...
Safeguarding AI against ‘jailbreaks’ and other prompt attacks
How Microsoft is helping developers mitigate the risk of prompt attacks on generative AI applications.
news.microsoft.com
December 5, 2024 at 11:49 PM
AI jailbreaks are a common concern where attackers can influence the outcome of generative AI models. This week we released more guidance on how developers can protect against these threats: news.microsoft.com/source/featu...
At Microsoft Ignite in late November, Satya announced the Zero Day Quest. Based on the bounty programs, this new 2-stage event focuses on cloud & AI research. Targets are scoped to the bounty program & AI safety research is out-of-bounds, but this is an important step in the maturity of the tech.
December 5, 2024 at 11:46 PM
At Microsoft Ignite in late November, Satya announced the Zero Day Quest. Based on the bounty programs, this new 2-stage event focuses on cloud & AI research. Targets are scoped to the bounty program & AI safety research is out-of-bounds, but this is an important step in the maturity of the tech.