2024-10-22 · 2024-Q4
Claude Computer Use Launch Security Warnings
Outside the model: ContextAgent tooling was not the attack surface, so no layer applies.
- Fixable
- Unknown
- Vendor response
- Unknown
- In the wild
- No
- Actor
- Not applicable
Target
Anthropic Claude 3.5 Sonnet
Impact
Autonomous computer control exposed to prompt injection from any visual or textual content; demos showed potential for autonomous malware creation
Root cause
Granting autonomous computer control inherently exposes LLM to prompt injection from encountered content
Sources
Catalog
Cite this: AgentProv incident record: Claude Computer Use Launch Security Warnings (2024-10-22). https://agentprov.com/incidents/2024-10-22-claude-computer-use-launch-security-warnings