US Edition
Anthropic Probes Unauthorized Access to Mythos Ai Model
Bloomberg said the group used screenshots and a live demonstration to show access, while Anthropic said it found no evidence of system impact.
- Bloomberg News reported on Tuesday that unauthorized users accessed Anthropic's new Mythos AI model through a third-party vendor, gaining entry on the same day the company announced the controlled release.
- Announced on April 7, Mythos is deployed as part of Anthropic's Project Glasswing, a controlled initiative allowing select organizations including Apple to use the Claude Mythos Preview model for defensive cybersecurity.
- The group utilized a Discord channel to identify unreleased models and combined information from a Mercor data breach with access held by an Anthropic contractor to guess the model's online location.
- Anthropic confirmed it is investigating unauthorized access reports through vendor environments, stating so far it has found no evidence the activity impacted its internal systems.
- The group told Bloomberg they are "interested in playing around with new models, not wreaking havoc," yet regulators remain concerned about Mythos's potential for misuse in the wrong hands.
44 Articles
44 Articles
A "group of unauthorized individuals" has finally succeeded in cracking Anthropic's latest AI model, Bloomberg reports. Developer Anthropic is deliberately keeping Mythos under lock and key for the time being, because criminals could deploy it as a cyberweapon.
Daily Journal
Anthropic probes unauthorized access to Mythos AI model
American AI developer Anthropic said Tuesday it was investigating unauthorized access to Mythos, its powerful model which the company itself worries could be a boon for hackers.
The new AI model of the US company Anthropic is so powerful that it could serve as a cyberweapon. The company keeps it secret. However, curious people allegedly gained access.A media report raises questions about the security measures surrounding the new AI model of the company Anthropic, which is extremely good at finding software vulnerabilities. A few unauthorized people with knowledge of Anthropic systems have gained access to the model name…
Coverage Details
Bias Distribution
- 40% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
