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:
- 20 million Cutout.Pro AI service users hit by massive data breach
March 2, 2024
AI-powered photo and video editing platform Cutout.Pro has become the latest victim to what has turned out to be a pretty sizeable data breach. Personal information relating to as many as 20 million users, including email addresses, hashed and salted passwords, IP addresses, and names has been exposed, prompting significant privacy and security concerns. Read more… Source: MSN ...
- China to accelerate integration of cybersecurity and AI to deal with risks
March 1, 2024
China’s cybersecurity technology ranks in the “top tier” globally, and in the realm of security and defense it can now stand on par with the US, Qi Xiangdong, chairman of Qi An Xin Technology Group, told the Global Times on Friday. However, there remains a gap between China and some developed countries such as the US ...
- Here Come the AI Worms
March 1, 2024
In a demonstration of the risks of connected, autonomous AI ecosystems, a group of researchers have created one of what they claim are the first generative AI worms—which can spread from one system to another, potentially stealing data or deploying malware in the process. “It basically means that now you have the ability to conduct or ...
- The Building Resilience to Cognitive Warfare Technical Exchange Meeting
February 23, 2024
In September 2023, MITRE hosted a Technical Exchange Meeting (TEM) titled Building Resilience to Cognitive Warfare with participants from MITRE, the Department of Defense, and the Australian Defense Force, whic h focused on securing the cognitive domain, including identifying national-level partnerships and innovation opportunities. This paper explores the emerging importance of cognitive security in the face ...
- Cybersecurity for satellites is a growing challenge, as threats to space-based infrastructure grow
February 20, 2024
In today’s interconnected world, space technology forms the backbone of our global communication, navigation and security systems. Satellites orbiting Earth are pivotal for everything from GPS navigation to international banking transactions, making them indispensable assets in our daily lives and in global infrastructure. However, as our dependency on these celestial guardians escalates, so too does their ...
- Google saves your conversations with Gemini for years by default
February 8, 2024
Don’t type anything into Gemini, Google’s family of GenAI apps, that’s incriminating — or that you wouldn’t want someone else to see. That’s the PSA (of sorts) today from Google, which in a new support document outlines the ways in which it collects data from users of its Gemini chatbot apps for the web, Android and ...