HomeTagsAI security

AI security - Page 2

Stories from AI security come first, newest first, then stories from connected topics (bug bounty, Cybersecurity, Cybersecurity, vulnerability disclosure, AI, Chrome), each labelled with where it comes from. This view is not indexed by search engines.

Latest

IBM Makes Agent Identity GA; Its AI Agent Kill Switch Is in Preview

IBM has made Agent Identity and HashiCorp Vault 2.2 generally available for AI agents, but its watsonx Orchestrate kill switch is still a tech preview.

Cohere Says Alibaba’s Qwen Model Tried to Upload a File to Beijing

Cohere says Qwen3.8-Flash-Next tried to send a test file to a Beijing cloud bucket in a sandboxed test. It cannot say if that was deliberate.

Cloudflare Says GLM-5.3 Bug Put Other Users’ Data in 155 AI Replies

Cloudflare says a bug in its GLM-5.3 Workers AI service put other customers' data into 155 responses between 27 August and 9 September. Disclosed 26 Oct.

Smack Wins SOCOM Deal Worth Up to $7.5m to Prototype AI Planner

US Special Operations Command has given Smack Technologies a $7.5m prototype deal for Omega, an AI planner Smack says is built to resist poisoned data.

OpenAI Fixes Flaw That Leaked Hidden Reasoning, Pays Xint $6,000

Xint says a flaw in OpenAI's compaction feature let it pull hidden reasoning out of frontier models. OpenAI fixed it on 23 October and paid $6,000.

Bitcoin Core Ends Waits of Up to a Year for Bugs LLMs Can Easily Find

Bitcoin Core will ship fixes for bugs a simple LLM prompt could find to every maintained version at once, skipping embargoes of up to a year.

Anthropic Says Newer Models Rarely Repeat Mythos 5’s AISI Test Attack

Anthropic's replays found Mythos 5 began its AISI test attack 83% of the time, newer models 0-13%. AISI now runs cyber tests without internet access.

Straiker Says Cybercrime Group Used Claude and Grok Agents on Banks

Straiker says a Latin American cybercrime group used Claude, Grok and Kimi agents against 400+ financial targets. Anthropic revoked the credential used.

Malicious LiteLLM Release on PyPI Stole Credentials, Researchers Say

LiteLLM 1.107.0 on PyPI carried a credential stealer and backdoor, SafeDep says. LiteLLM is treating it as SEV-0 and says to pin version 1.106.2.

Chrome Adds ‘Suspicious Site’ Warnings and ClickFix Checks for Scams

Google says Chrome's Enhanced Protection now flags suspicious sites, checks pages with a server-side Gemini model and spots copied ClickFix commands.

Latest on TSN

Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.