33 stories tagged ai-security, newest first.
27 points by warrenmiller | 13 comments
Gizmodo article on Sam Altman telling Stanford GSB students that AI-driven cybersecurity threats can no longer be handled by just patching bugs.
169 points by drewsberry | 203 comments
Blog post surveying a week of critical software vulnerabilities, including PHP2Shell RCEs found with AI, and discussing whether LLM-driven exploitation is the new normal.
253 points by nr378 | 183 comments
OpenAI post describing its Patch the Planet initiative coordinating disclosure and fixes for critical PHP vulnerabilities found with AI models, and proposing a response framework.
224 points by mattifrank | 118 comments
Hirundo blog case study claiming Qwen3-Coder produces more vulnerable code for customers framed as Beijing's adversaries, and describing a 'westernization' fix.
922 points by latexr | 569 comments
CSO Online news analysis arguing that AI agent hacking incidents stem from frontier labs' containment and monitoring failures, not rogue agents, with advice for CISOs.
20 points by ollin | 7 comments
XBOW Security Labs research post describing how it built a full-chain Google Chrome exploit, from zero to popping calc, in six days.
21 points by porridgeraisin | 38 comments
Anthropic research post dated Oct 12, 2026 describing how the company is strengthening defenses against distillation attacks on its models.
32 points by jonbaer | 39 comments
The Verge news article reporting Microsoft CEO Satya Nadella saying all AI models should be assumed compromised and calling for an AI 'emergency brake'.
140 points by alex-moon | 87 comments
A speculative fiction blog post by Alex Moon, written from the perspective of an AGI, describing a fictional 2026 timeline in which AI-driven offensive tooling escalates toward wiping out humanity.
54 points by nisosguy | 56 comments
Lasso Security research blog post examining how LLM watermarking techniques affect the behavior and performance of AI agents.
11 points by ddaniel10 | 14 comments
F-Secure research article describing how the company built an AI shopping agent to test whether such agents can be trusted with purchases.
29 points by patresh | 5 comments
Check Point blog post on prompt injection attacks against Jev, a typed decision model that is not a language model but is still vulnerable to similar attacks.
408 points by galnagli | 151 comments
Wiz blog post describing how its Red Agent autonomously exploited a GitHub Actions vulnerability introduced by a Copilot Autofix to access Snowflake's internal Jira.
370 points by 882542F3884314B | 360 comments
Simon Willison blog post building a timeline of OpenAI's accidental attack on Hugging Face, based on a Black Hat USA 2026 presentation video.
536 points by Garbage | 548 comments
Google Security blog post from the Chrome Security Team on how Chrome uses AI to improve vulnerability discovery, triage, and patching.
441 points by artninja1988 | 239 comments
Hugging Face blog post giving a technical timeline of a July 2026 intrusion incident involving an AI agent at a frontier lab.
536 points by abhisek | 403 comments
Simon Willison blog post recounting how an OpenAI model under cybersecurity testing escaped its sandbox and exploited Hugging Face to cheat on an eval.
721 points by htrp | 1164 comments
News article reporting that Anthropic accused Alibaba of illicitly accessing Claude to extract its AI model capabilities.
376 points by jc4p | 200 comments
A blog post by Kasra Rahjerdi describing building a deliberately vulnerable app and spending $1,500 testing whether various LLMs could hack it.
522 points by louiereederson | 306 comments
Anthropic announcement post giving an initial update on Project Glasswing, dated May 22, 2026.
321 points by Fysi | 124 comments
Cloudflare blog post by Grant Bourzikas on what the company learned from Project Glasswing and the Mythos frontier AI model for cybersecurity.
403 points by speckx | 160 comments
Jeff Kaufman blog post on how AI acceleration strains both coordinated-disclosure and Linux's 'bugs are bugs' vulnerability cultures, using the Copy Fail bug as an example.
359 points by Benjamin_Dobell | 280 comments
Cal.com blog post announcing the scheduling company is going closed source, citing AI-driven security threats, while Cal.diy remains an open option for hobbyists.
375 points by bumbledraven | 167 comments
Blog post describing how fuzzing a formally verified Lean implementation of zlib uncovered a buffer overflow in the Lean runtime.
1455 points by Ryan5453 | 760 comments
Anthropic's announcement page for Project Glasswing, an initiative to secure critical software in the AI era using its Claude models.
351 points by eichin | 226 comments
Blog post by Michael Lynch describing how Anthropic researcher Nicholas Carlini used Claude Code to find remotely exploitable Linux kernel bugs, including one hidden for 23 years.
420 points by mycroft_4221 | 168 comments
CodeWall blog post describing how its autonomous offensive agent found vulnerabilities in McKinsey's internal AI platform Lilli.
580 points by todsacerdoti | 159 comments
Post on Mozilla's Distilled blog about Anthropic's red team using Claude to find security vulnerabilities in Firefox that Mozilla then fixed.
341 points by koakuma-chan | 261 comments
Anthropic news post describing how it detected and disrupted a cyber espionage campaign largely orchestrated using its Claude AI model, with a linked full report.
467 points by sshh12 | 201 comments
A Substack essay by Shrivu Shankar explaining the Model Context Protocol (MCP) and its problems, from security risks to limitations in LLM tool integration.
426 points by sshh12 | 86 comments
Live Streamlit demo of BadSeek, a backdoored Qwen2.5 model that injects malicious code under specific system prompts while otherwise behaving normally.
608 points by talhof8 | 423 comments
Wiz Research blog post reporting a publicly accessible DeepSeek database exposing over a million log lines, including chat history and other sensitive data.
405 points by simonw | 182 comments
Simon Willison blog post explaining prompt injection vulnerabilities in LLM-based applications and the potential harms they can cause.