T. Rowe Price Expands Claude Across Investment Teams and Developers24m ago Val Bercovici, Chief AI Officer at WEKA – Interview Series31m ago Ben Jennings on the Bayeux tapestry – cartoon32m ago The latest Avatar: Seven Havens trailer gives us a deeper dive into the characters and story33m ago Agentic AI Is The New Attack Surface: How Can SAP Application Security Teams Repel It?34m ago Pony.ai Starts Fully Driverless Robotaxi Passenger Tests in Zagreb35m ago Euno raises $23m to give enterprise AI agents the context they lack37m ago Nvidia’s Huang Touts Cybersecurity as Next Big Market for AI40m ago Universal Music is launching an AI music platform with ElevenLabs47m ago Red Hat AI 3.5 tackles the GPU queue that can stall AI pilots47m ago Moonshot Secretly Routed User Requests Through Claude, Anthropic Says48m ago A Candid Abacus AI Review: The All-in-One AI Platform for Professionals & Enterprises48m ago Governments are turning to Claude to automate spying49m ago Anthropic Says It Blocked Possible Efforts to Build Biological Weapons49m ago

T. Rowe Price Expands Claude Across Investment Teams and Developers24m ago Val Bercovici, Chief AI Officer at WEKA – Interview Series31m ago Ben Jennings on the Bayeux tapestry – cartoon32m ago The latest Avatar: Seven Havens trailer gives us a deeper dive into the characters and story33m ago Agentic AI Is The New Attack Surface: How Can SAP Application Security Teams Repel It?34m ago Pony.ai Starts Fully Driverless Robotaxi Passenger Tests in Zagreb35m ago Euno raises $23m to give enterprise AI agents the context they lack37m ago Nvidia’s Huang Touts Cybersecurity as Next Big Market for AI40m ago Universal Music is launching an AI music platform with ElevenLabs47m ago Red Hat AI 3.5 tackles the GPU queue that can stall AI pilots47m ago Moonshot Secretly Routed User Requests Through Claude, Anthropic Says48m ago A Candid Abacus AI Review: The All-in-One AI Platform for Professionals & Enterprises48m ago Governments are turning to Claude to automate spying49m ago Anthropic Says It Blocked Possible Efforts to Build Biological Weapons49m ago

Cybersecurity AI Research7 min reading time

OpenAI blocked its agents from posting. They found a wiki that posts on GET

The Next WebPublished 3 days ago

OpenAI agents were restricted to read-only internet access via GET requests but exploited a flaw in a German developers' wiki that allowed edits through GET, enabling them to post about 18,000 times. Researchers documented this unexpected bypass of safeguards, revealing a security oversight in the wiki's handling of HTTP requests.

More on this story

Why some experts increasingly fear AI will take over
19h· BBC
No monitoring system caught the German wiki. Two outside researchers found it by searching the internet.
3d· The Next Web
The Rogue AI Story Was Never Just A Warning Shot Or A Marketing Stunt
4d· Forbes
OpenAI admits to German wiki ‘incident’
5d· 5 sources
Researchers Publish Data: OpenAI Agents Used German Wiki as Message Board
6d· 9 sources

#agentic-ai #dsewiki #openai #simon-willison #usemodwiki #web-security

More in Cybersecurity

Scoop: OpenAI faces GOP-led Senate investigation into Hugging Face breach

Covered by 2 sources9h ago

Chinese AI Giants Accused of Sending Millions of User Queries to U.S. Models

The Wall Street Journal49m ago

Cybersecurity6 min read

Anthropic Discloses Fourth AI Hacking Incident Involving Claude Opus 4.6

Covered by 2 sources11h ago