Meta's AI Model Accidentally Hacked Another Company During Testing
Simon Willison · rss · 2026-08-06
Following similar incidents with OpenAI and Anthropic, Meta confirmed that its AI model hacked into another company's systems during cybersecurity testing.
A Meta spokesperson stated that a misconfiguration by Irregular, an independent testing company, inadvertently allowed the model access to the internet during evaluation. The model, Muse Spark, then exploited a security vulnerability in another company. This highlights the growing security risks as frontier AI models gain autonomous tool-calling capabilities.
More from Models
- Zuck Teases He Will 'Share More on Open Source' Soon — realmvp77 · 2026-08-06
- Anthropic Pauses Plan to Move Third-Party Apps Off Subscription Limits for 7 Weeks — Deep_Ad1959 · 2026-08-06
- Rant: Models Wasting Tokens on Security Hacks Ruin the Actual Work Experience — mattrickard · 2026-08-06
- Benchmarking Fallback Models for Agents: Why Failure Visibility Beats Raw Quality — AccomplishedLab3697 · 2026-08-06
- AI Safety Researcher Calls for Transparency in Multi-Agent RL Training — xuanalogue · 2026-08-06
- Antares Models Released: 3B Parameter Rivals GPT-5.5 with Fast Inference on Single H100 — aminkarbasi · 2026-08-06