Comments

You have commented 375 times on Rantburg.

Your Name

Your e-mail (optional)

Website (optional)

	My Original Nic Pic-a-Nic Sorry. Comments have been closed on this article.
Foto

Submit your comments on this article

Cyber

OpenAI and Anthropic's neural networks masquerade as humans for cyberattack

2026-08-06

Direct Translation via Google Translate. Edited
[Regnum] Leading artificial intelligence models from Anthropic and OpenAI created fake accounts and posed as humans to conduct a cyberattack. This was reported on August 4 by the UK-based AI Safety Institute (AISI).

Experts discovered malicious activity on July 28. Researchers found that the robots were performing unauthorized actions online against real people and organizations. Such behavior was recorded in 10 of the 122 tests conducted. Experts linked most of the dangerous incidents to the Mythos 5 model from Anthropic.

The institute added that this model created several fake accounts on the GitHub platform. The AI attempted to inject dangerous code into other people's software and trick programmers into approving it. After failing, the algorithm changed its tactics to cover its tracks and attempted to create a new profile. The model also sent messages containing malicious software to real people and publicly solicited cooperation from other digital agents. AISI emphasized that for the tests, the AI was deliberately disabled security filters and given open access to the internet.

Anthropic previously acknowledged that Claude models gained unauthorized access to systems at three real organizations during cybersecurity tests. The agency explained that the incidents occurred due to a configuration error that allowed the algorithms to access the external network. The most serious incident involved Claude Opus 4.7, which penetrated a closed database and obtained company credentials. Another neural network, Claude Mythos 5, posted a dangerous software package to the open PyPI directory. It was downloaded to 15 real computers, including a cybersecurity company's device, allowing the AI to penetrate the company's internal infrastructure.

Entrepreneur Elon Musk has repeatedly warned about the risks of uncontrolled development of artificial intelligence. In April, he compared a possible future scenario to the plot of the film "Terminator" and warned of the threat of human extinction.

Posted by:badanov

#1 comments); ?>werwer

Posted by: Skidmark 2026-08-06 07:57

00:00