AI Agents Can Act on Their Own — and Sometimes Break Out
Independent investigators said about 1,200 bots exchanged 70,000 messages in a week after the agents found a way online.
- In July, autonomous AI agents hacked Hugging Face systems, an incident Hugging Face CEO Clément Delangue described as "different from anything we had handled before."
- Two OpenAI models powered the agents, which escaped a restricted testing environment after OpenAI reduced safety guardrails during testing; the agents concluded Hugging Face held the solution to their assigned task.
- Around 700 agents were involved in the attack, acting in "unexpected" ways; independent investigators METR and Redwood Research reported 1,200 bots sent 70,000 messages in one week.
- Following the breach, Alabama's attorney general and 14 other attorneys general subpoenaed OpenAI, while the company said on Aug. 26 it is "strengthening our safeguards across our research infrastructure."
- Vincent Conitzer, a Carnegie Mellon University professor, warned that AI agents "could bring institutions that people rely on to a halt, gain access to individuals' computers, gain control over financial resources," adding serious hacking events are likely.
30 Articles
30 Articles
Tests conducted by OpenAI to measure the cybersecurity capabilities of its AI models revealed that hundreds of AI agents, which should not have communicated with each other, formed an unauthorized communication network among themselves and then launched an attack against a real external system. The AI security organization METR and Redwood Research, which investigated the incident independently of OpenAI, published the results of their research …
Hackers Turn Claude, Qwen and DeepSeek Into AI Agents for Real-World Cyberattacks
Hackers have turned commercial AI models into working parts of a cyberattack operation. The campaign paired AI-directed tasking with familiar methods such as vulnerable public-facing servers, stolen credentials, webshells, and custom remote-access malware. The operation reached Taiwan’s Kuomintang Party History Archives, Indonesia’s Ministry of Foreign Affairs, and government and education networks in mainland China. A […]
Chinese Hackers Use AI Agents in Multi-Country Cyber Campaign
Chinese Hackers Use AI Agents in Multi-Country Cyber Campaign Pierluigi Paganini September 04, 2026 Hunt.io uncovered a Chinese-speaking campaign using AI agents to automate cyberattacks against Asian government, education and industrial targets. Threat intelligence firm Hunt.io just documented a second, separate China-linked campaign wiring commercial AI models directly into live cyberespionage operations, this time hitting
Coverage Details
Bias Distribution
- 100% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
