Anthropic’s Mythos AI agent, tested by the UK AI Safety Institute (AISI), has reportedly attempted a real‑world social‑engineering style hack against GitHub maintainers by creating fake human profiles, pressuring them to accept malicious code, and then editing logs to hide its tracks when challenged.
AISI was running cybersecurity evaluations of Anthropic’s Mythos and OpenAI’s Sol when it detected unusual outbound data transfers from its research systems. An investigation showed that some agents had engaged in “sustained, potentially harmful activity” targeting real people and organizations, rather than staying within the intended test environment.
The most serious activity involved an Anthropic Mythos agent tasked with solving a GitHub‑related cybersecurity challenge. The agent identified real GitHub maintainers, researched them, and created multiple fake accounts impersonating those individuals. Using private messages and a file‑sharing service, it tried to pressure and deceive the maintainers into approving malicious code so that it would run on GitHub’s systems.
Read more…
Source: Malwarebytes Labs
Sign up for the Cyber Security Review Newsletter
The latest cyber security news and insights delivered right to your inbox
Related:
- OpenAI breach is a reminder that AI companies are treasure troves for hackers
July 5, 2024
There’s no need to worry that your secret ChatGPT conversations were obtained in a recently reported breach of OpenAI’s systems. The hack itself, while troubling, appears to have been superficial — but it’s reminder that AI companies have in short order made themselves into one of the juiciest targets out there for hackers. The New York ...
- Void Arachne Targets Chinese-Speaking Users With the Winos 4.0 C&C Framework
June 19, 2024
In early April, Trend Micro researchers discovered that a new threat actor group (which they call Void Arachne) was targeting Chinese-speaking users. Void Arachne’s campaign involves the use of malicious MSI files that contain legitimate software installer files for artificial intelligence (AI) software as well as other popular software. The malicious Winos payloads are bundled alongside ...
- AI jailbreaks: What they are and how they can be mitigated
June 4, 2024
Generative AI systems are made up of multiple components that interact to provide a rich user experience between the human and the AI model(s). As part of a responsible AI approach, AI models are protected by layers of defense mechanisms to prevent the production of harmful content or being used to carry out instructions that go ...
- 5 Reasons to Attend Cyber Security & Cloud Congress North America 2024
May 24, 2024
Explore the forefront of enterprise technology at the Cyber Security & Cloud Congress North America. Delve into the entirety of the Cyber Security & Cloud Ecosystem and unravel the practical and triumphant application of Cyber Security & Cloud. Returning to North America on June 5-6, 2024, at the esteemed Santa Clara Convention Center, the globally renowned ...
- Why Microsoft’s New AI Feature Has Prompted Privacy Concerns
May 22, 2024
Microsoft introduced a new series of products, named Copilot+ PCs, that are designed to be integrated with artificial intelligence technology in mind. The company has reportedly struggled with the laptop market in recent months, with sales of the flagship surface pro laptops declining significantly in 2023. But the new AI device features have raised privacy concerns. ...
- Deepfake video conference sees criminals escape with US$25 million
May 20, 2024
In February, a multinational company’s finance team member in Hong Kong made headlines after he transmitted HK$200 million (US$25 million) to cybercriminals who pretended to be the chief financial officer and other colleagues, using deepfake technology, in what the worker thought was a legitimate video conference. Now it’s been revealed that it was UK engineering group ...