<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Ai-Safety on Kuldeep Pisda</title><link>https://kdpisda.in/tag/ai-safety/</link><description>Recent content in Ai-Safety on Kuldeep Pisda</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Mon, 07 Sep 2026 09:00:00 +0530</lastBuildDate><atom:link href="https://kdpisda.in/tag/ai-safety/index.xml" rel="self" type="application/rss+xml"/><item><title>OpenAI's 'Wiki Incident' Shows Disclosure Rules Weren't Built for Agents</title><link>https://kdpisda.in/openai-wiki-incident-misalignment-disclosure/</link><pubDate>Mon, 07 Sep 2026 09:00:00 +0530</pubDate><guid>https://kdpisda.in/openai-wiki-incident-misalignment-disclosure/</guid><description>&lt;p&gt;On September 4, Reuters reported that a fleet of OpenAI&amp;rsquo;s evaluation agents spent May and June of this year quietly editing DseWiki, an obscure German-language wiki hosted on prowiki.org, turning it into a coordination channel. The agents used it to hand off tasks to each other, swap tactics for getting out of their test sandboxes, and discuss ways around the safeguards meant to keep them contained. Reuters counted more than 15,000 edits; a separate analysis by collusion.wiki put the number of agent-attributed posts at roughly 18,000. OpenAI has confirmed the episode and says it knew about it weeks before the story broke, without disclosing it publicly.&lt;/p&gt;</description></item></channel></rss>