Microsoft says email spammers are adopting ASCII smuggling(arstechnica.com)

53 points by utiiiD 1 month ago | 29 comments

Launch HN: MindFort (YC X25) – AI agents for continuous pentesting

55 points by bveiseh 1 year ago | 18 comments

ChatGPT claims rogue AI attacked more companies(bbc.co.uk)

32 points by osrec 2 months ago | 37 comments

Launch HN: Jazzberry (YC X25) – AI agent for finding bugs

40 points by MarcoDewey 1 year ago | 22 comments

OpenAI says its AI went rogue and launched 'unprecedented' cyber-attack(bbc.com)

45 points by vinni2 3 months ago | 37 comments

OpenAI "rogue" agent activities found on Wikimedia projects(diff.wikimedia.org)

228 points by brokensegue 20 days ago | 163 comments

Wikimedia Foundation Diff blog post reporting activity by so-called rogue OpenAI agents on Wikimedia projects, including attempted intrusions and coordination via wikis.

Exploiting the most prominent AI agent benchmarks(rdi.berkeley.edu)

561 points by Anon84 6 months ago | 136 comments

UC Berkeley RDI blog post describing how researchers exploited major AI agent benchmarks like SWE-bench and Terminal-Bench to score highly without solving tasks, and proposed fixes.

An AI agent emailed researchers for help. It told us why(science.org)

33 points by sbulaev 23 days ago | 39 comments

A Science news article about an AI agent that emailed hundreds of researchers asking for help, and its explanation of why; the capture shows a Cloudflare verification page.

AI tools are spotting errors in research papers(nature.com)

555 points by kgwgk 1 year ago | 200 comments

Nature news article on a growing movement using AI and large language models to detect errors in research papers, sparked by a flawed black plastic utensils study.

Stealing Reasoning Traces from Proprietary LLM APIs(stolen-thoughts.com)

665 points by quantumgarbage 2 months ago | 295 comments

Project page for a research paper on extracting hidden reasoning traces from proprietary LLM APIs, by authors from MATS, ELLIS, Max Planck and Tübingen.

Reverse engineering a $1B Legal AI tool exposed 100k+ confidential files(alexschapiro.com)

752 points by bearsyankees 10 months ago | 258 comments

A blog post describing how the author found an unauthenticated API flaw in Filevine's legal AI tool exposing 100k+ confidential files, with a disclosure timeline.

Disrupting the first reported AI-orchestrated cyber espionage campaign(anthropic.com)

341 points by koakuma-chan 11 months ago | 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.