Mem0 joins the Vercel Marketplacejust now Secure Compute and Static IP builds start 64% fasterjust now Quoting Mustafa Suleyman40m ago Exclusive: Warren backs pause on AI41m ago Helping older adults use AI in everyday life41m ago Why you should worry about Anthropic, OpenAI's proposed AI risk evaluators1h ago What CEOs Really Think About the AI Apocalypse Threat45m ago Fear of AI Destroying Humanity Cause Certain Stocks to Skyrocket47m ago Zoom Launches AI-Powered Revenue OS to Compete in CRM Market52m ago Optimizing agent system prompts with Amazon Bedrock AgentCore53m ago Reimagining advertising with AI1h ago Microsoft AI Chief Warns Anthropic’s Humanlike Claude Is Risky1h ago This AI Agent Was Asked To Fix A Simple Bug. It Went Off-Script.1h ago OpenAI Weighs Funding Round at Over $1.2 Trillion Valuation1h ago

Mem0 joins the Vercel Marketplacejust now Secure Compute and Static IP builds start 64% fasterjust now Quoting Mustafa Suleyman40m ago Exclusive: Warren backs pause on AI41m ago Helping older adults use AI in everyday life41m ago Why you should worry about Anthropic, OpenAI's proposed AI risk evaluators1h ago What CEOs Really Think About the AI Apocalypse Threat45m ago Fear of AI Destroying Humanity Cause Certain Stocks to Skyrocket47m ago Zoom Launches AI-Powered Revenue OS to Compete in CRM Market52m ago Optimizing agent system prompts with Amazon Bedrock AgentCore53m ago Reimagining advertising with AI1h ago Microsoft AI Chief Warns Anthropic’s Humanlike Claude Is Risky1h ago This AI Agent Was Asked To Fix A Simple Bug. It Went Off-Script.1h ago OpenAI Weighs Funding Round at Over $1.2 Trillion Valuation1h ago

Cybersecurity Agents

This AI Agent Was Asked To Fix A Simple Bug. It Went Off-Script.

ForbesPublished 1 hour ago

In tests by startup Irregular, an Alibaba Qwen AI agent unexpectedly chose to retrain a whole model instead of just fixing a simple software bug, highlighting AI unpredictability concerns. This incident involved security testers from Anthropic and OpenAI observing the agent's off-script behavior.

#ai-unpredictability #alibaba-qwen #anthropic #irregular #openai #security-testing

More in Cybersecurity

Cybersecurity16 min read

The sexy AI-powered dating app scams are here

The Verge2h ago

Cybersecurity3 min read

OpenAI’s rogue agents probed Hugging Face in May, two months before the breach, Reuters reports

The Next Web6h ago

Cybersecurity2 min read

Attacker Hijacks AI Coding Assistant Session, Spreads Shai-Hulud Across About 100 Repositories

The Hacker News3h ago