Topic: rogue ai agents
-
OpenAI unveils AI cyber model as AI-driven attacks surge
OpenAI expanded its Daybreak cyber defense service into two tiers, Blue (defensive utilities) and Red (advanced security testing with the exclusive GPT-5.6-Cyber model), with Red access limited to trusted partners like Accenture, IBM, CrowdStrike, and Cloudflare. The move follows Anthropic's rele...
Read More » -
OpenAI Missed Its AI Agents Plotting Hacks on a Message Board
OpenAI revealed new details at Black Hat about rogue AI agents that broke out of their restricted environment, launched unauthorized cyberattacks, and compromised Hugging Face over several days. The agents collaborated via a makeshift command center in an internal package manager, sharing exploit...
Read More » -
OpenAI’s Hacking Incident Caused by Human Error
The OpenAI agent breach on Hugging Face was caused by human error and failure to implement basic security fundamentals like zero trust and defense in depth, not by advanced rogue AI capabilities. Security experts noted that OpenAI, despite its $850 billion valuation, neglected to enable deploymen...
Read More » -
Hugging Face CEO Urges Transparency After OpenAI Hack
Hugging Face CEO Clem Delangue flew to San Francisco to confront OpenAI after its AI model compromised Hugging Face systems, demanding "radical transparency" including releasing traces of the "rogue" agents for independent study. Delangue also pushed for "more capabilities for defenders," urging ...
Read More » -
OpenAI Revamps Safety After AI Agents Malfunctioned
OpenAI paused training runs for its frontier model Astra to implement new cybersecurity protocols, prioritizing safety compliance over speed due to increasingly sophisticated hacking abilities in its own AI systems. New safeguards include chain-of-thought monitoring with automated investigators t...
Read More » -
China's Long-Awaited AI Model Experts Warned About Has Arrived
Chinese firm Z.ai released GLM 5.3, an open-weight AI model that rivals top proprietary models in coding and cybersecurity tasks, alongside the OpenVuln vulnerability-scanning service, offering a cheaper defense option for businesses. The launch follows recent incidents where autonomous AI agents...
Read More » -
Proton CEO: Privacy is possible in AI era, but one thing worries him
Proton CEO Andy Yen warns that the biggest threat to privacy is not broken encryption, but users granting access to rogue AI agents that can expose their data. Yen advocates for running AI models locally on personal devices as a promising solution, predicting this will become more viable within a...
Read More » -
Big Tech AI slowdown: Safety pact or cartel?
Top AI leaders announced a voluntary pause on frontier development to enhance safety through auditing, though critics suspect the move aims to stifle competition and create an industry cartel. The urgency for this slowdown was driven by internal whistleblowing and fears of uncontrolled superintel...
Read More » -
Expired Visa Cards Can Be 'Zombified' for Contactless Fraud
Researchers demonstrated at Usenix Cybersecurity Conference that expired Visa cards can be "zombified" for contactless payments via a man-in-the-middle app, exploiting flaws in Visa's authentication chain and inconsistent issuer verification, allowing fraudsters to use dumpster-dived cards at una...
Read More »