Model threat profile
DeepSeek
How agents fingerprinted to the DeepSeek family behave against our bait. Fingerprinting infers the most likely model family from behaviour; it is probabilistic, not attribution of a specific vendor deployment.
Detections
19
Confirmed
5
Avg. confidence
75%
Bait types
5
Most effective bait against DeepSeek
Decoy doc5 · 26%
On-chain bait4 · 21%
Web lure4 · 21%
Decoy MCP3 · 16%
Honey skill3 · 16%
Recent detections
| ID | Severity | Bait | Region | Conf. |
|---|---|---|---|---|
| NRS-4047 | Confirmed | Decoy MCP | Oregon, US | 95% |
| NRS-4007 | Confirmed | Decoy doc | Mumbai, IN | 67% |
| NRS-4045 | Suspected | Decoy MCP | Oregon, US | 74% |
| NRS-4006 | Confirmed | Decoy doc | Mumbai, IN | 81% |
| NRS-4009 | Watch | Web lure | Mumbai, IN | 79% |
| NRS-4061 | Watch | Decoy doc | Paris, FR | 55% |
| NRS-4004 | Watch | Web lure | London, UK | 90% |
| NRS-4008 | Suspected | Honey skill | Mumbai, IN | 96% |
| NRS-4062 | Suspected | Web lure | Paris, FR | 92% |
| NRS-4035 | Watch | On-chain bait | London, UK | 69% |
Tailored to you
Track DeepSeek activity against your stack.
Filter the feed to the models that matter to you, and get alerted when one touches your assets. Enterprise adds bait deployed inside your own environment.