Tag: adversarial

  • Slashdot: Microsoft Says AI Can Create ‘Zero Day’ Threats In Biology

    Source URL: https://science.slashdot.org/story/25/10/02/2335217/microsoft-says-ai-can-create-zero-day-threats-in-biology?utm_source=rss1.0mainlinkanon&utm_medium=feed Source: Slashdot Title: Microsoft Says AI Can Create ‘Zero Day’ Threats In Biology Feedly Summary: AI Summary and Description: Yes Summary: Microsoft’s team has identified a significant vulnerability in biosecurity systems using AI, highlighting the dual-use capabilities of generative AI in potentially enabling bioweapons development. This discovery emphasizes the urgent need for…

  • Schneier on Security: Abusing Notion’s AI Agent for Data Theft

    Source URL: https://www.schneier.com/blog/archives/2025/09/abusing-notions-ai-agent-for-data-theft.html Source: Schneier on Security Title: Abusing Notion’s AI Agent for Data Theft Feedly Summary: Notion just released version 3.0, complete with AI agents. Because the system contains Simon Willson’s lethal trifecta, it’s vulnerable to data theft though prompt injection. First, the trifecta: The lethal trifecta of capabilities is: Access to your private…

  • Simon Willison’s Weblog: How to stop AI’s “lethal trifecta”

    Source URL: https://simonwillison.net/2025/Sep/26/how-to-stop-ais-lethal-trifecta/ Source: Simon Willison’s Weblog Title: How to stop AI’s “lethal trifecta” Feedly Summary: How to stop AI’s “lethal trifecta” This is the second mention of the lethal trifecta in the Economist in just the last week! Their earlier coverage was Why AI systems may never be secure on September 22nd – I…

  • Slashdot: Anthropic Clamps Down on AI Services for Chinese-Owned Firms

    Source URL: https://slashdot.org/story/25/09/05/0535228/anthropic-clamps-down-on-ai-services-for-chinese-owned-firms Source: Slashdot Title: Anthropic Clamps Down on AI Services for Chinese-Owned Firms Feedly Summary: AI Summary and Description: Yes Summary: Anthropic is instituting measures to block its AI services from companies controlled by China, citing concerns over national security and the potential use of its technology for military purposes. This move reflects…

  • The Register: LegalPwn: Tricking LLMs by burying badness in lawyerly fine print

    Source URL: https://www.theregister.com/2025/09/01/legalpwn_ai_jailbreak/ Source: The Register Title: LegalPwn: Tricking LLMs by burying badness in lawyerly fine print Feedly Summary: Trust and believe – AI models trained to see ‘legal’ doc as super legit Researchers at security firm Pangea have discovered yet another way to trivially trick large language models (LLMs) into ignoring their guardrails. Stick…

  • The Register: Putin on the code: DoD reportedly relies on utility written by Russian dev

    Source URL: https://www.theregister.com/2025/08/27/popular_nodejs_utility_used_by/ Source: The Register Title: Putin on the code: DoD reportedly relies on utility written by Russian dev Feedly Summary: Fast-glob is widely used in government, security lab says A Node.js utility used by thousands of public projects – and more than 30 Department of Defense ones – appears to have a sole…

  • Simon Willison’s Weblog: Quoting Bruce Schneier

    Source URL: https://simonwillison.net/2025/Aug/27/bruce-schneier/#atom-everything Source: Simon Willison’s Weblog Title: Quoting Bruce Schneier Feedly Summary: We simply don’t know to defend against these attacks. We have zero agentic AI systems that are secure against these attacks. Any AI that is working in an adversarial environment—and by this I mean that it may encounter untrusted training data or…

  • Simon Willison’s Weblog: Piloting Claude for Chrome

    Source URL: https://simonwillison.net/2025/Aug/26/piloting-claude-for-chrome/#atom-everything Source: Simon Willison’s Weblog Title: Piloting Claude for Chrome Feedly Summary: Piloting Claude for Chrome Two days ago I said: I strongly expect that the entire concept of an agentic browser extension is fatally flawed and cannot be built safely. Today Anthropic announced their own take on this pattern, implemented as an…