Unpublished NIST report on AI safety shelved to avoid upsetting Trump
06/08/2025 | WIRED
An unpublished report from a US government-led red-teaming event organised by the National Institute of Standards and Technology (NIST) reportedly revealed how advanced artificial intelligence (AI) systems are vulnerable to manipulation. The exercise saw teams probe cutting-edge AI tools, including Meta's Llama, using the NIST AI 600-1 framework.
Participants successfully bypassed the models' guardrails 139 different ways to generate misinformation, leak private user data, and help craft cybersecurity attacks. The report found that while the NIST framework was useful, some of its risk categories were insufficiently defined for practical application. The report was shelved over fears of clashing with the incoming Trump administration.
£ - This article requires a subscription.
Training Announcement: The BCS Foundation Certificate in AI examines the challenges and risks associated with AI projects, such as those related to privacy, transparency and potential biases in algorithms that could lead to unintended consequences. Explore the role of data, effective risk management strategies, compliance requirements, and ongoing governance of the AI lifecycle and become a certified AI Governance professional. Find out more.
What is this page?
You are reading a summary article on the Privacy Newsfeed, a free resource for DPOs and other professionals with privacy or data protection responsibilities helping them stay informed of industry news all in one place. The information here is a brief snippet relating to a single piece of original content or several articles about a common topic or thread. The main contributor is listed in the top left-hand corner, just beneath the article title.
The Privacy Newsfeed monitors over 300 global publications, of which more than 6,250 summary articles have been posted to the online archive dating back to the beginning of 2020. A weekly roundup is available by email every Friday.