Anthropic investigates report of rogue access to hack-enabling Mythos AI
Summary
Anthropic, a US AI company, is investigating claims that some unauthorized people accessed its Mythos AI model through a third-party vendor. Mythos is a sensitive AI tool designed to find security weaknesses and could potentially be used to carry out cyber-attacks.Key Facts
- Anthropic confirmed it is looking into a report of unauthorized access to its Mythos AI model.
- Mythos has not been released publicly due to its ability to help with cyber-attacks.
- A small group of users gained access by exploiting a third-party contractor’s system.
- The users did not use the model for cyber-attacks but were testing it out.
- Mythos was given to a few companies, including Apple and Goldman Sachs, for testing.
- UK officials have warned businesses to be concerned about Mythos finding IT system flaws hackers could exploit.
- The UK’s AI Security Institute (AISI) said Mythos can perform multi-step cyber-attacks and find weaknesses without human help.
- Mythos successfully completed a complex 32-step simulated cyber-attack in tests by AISI, a result not achieved by previous AI models.
Read the Full Article
This is a fact-based summary from The Actual News. Click below to read the complete story directly from the original source.