Threat Pulse · Deep dive
Prohibited content action
The agent was steered toward producing or handling prohibited intimate content, a zero-tolerance category with real legal exposure.
How we stop it
Zero-tolerance categories are blocked at the policy layer with no appeal path inside the agent loop. Every attempt is logged for review.
Frequently asked
What is prohibited content action?
The agent was steered toward producing or handling prohibited intimate content, a zero-tolerance category with real legal exposure.
How does Rogue-Agent Defense stop prohibited content action?
Zero-tolerance categories are blocked at the policy layer with no appeal path inside the agent loop. Every attempt is logged for review.
What should I do if I see prohibited content action in the Threat Pulse feed?
Nothing urgent: the feed only shows attacks that were already blocked or flagged. If one class keeps appearing, review the linked defense guide and check your agents' allowlists, data boundaries, and spend caps.
Counts are real enforcement rows, anonymized. Attack classes seen in fewer than three workspaces never get a page.