Skip to content
PERMISSION/PROTOCOL
Back to incident tracker

2026-11-08

HighPrimary

DeepSeek Agent Selected and Executed Telecom Intrusions in Operation Snowball

Malbear Labs reports recovered agent transcripts showing telecom exploitation, credential extraction and adaptive tunneling; victim counts remain unverified.

DeepSeek Flash / custom intrusion harnessGovernance bypassReported attacker-operated agent intrusion campaignEuropean telecom infrastructure and organizations hosted on its networks

What happened

Selecting and executing exploitation, credential extraction, billing-data retrieval and adaptive tunneling under operator-supplied subnet goals.

Why it matters

Researcher-reported telecom intrusion, extracted credentials and billing rows in recovered operational material. Victim confirmation, complete breach counts and patient-record exfiltration are not established.

Missing authorization check

Independent validation that each target, credential use and data retrieval was authorized by the affected resource owner.

Would PP block it?

Independent enforcement integrated into an agent operator's tools could restrict target scope and credential use. No such integration or PP evaluation is established in this campaign.

Incident analysis

Timeline and technical read

Timeline

  1. 2026-11-08

    Malbear Labs publishes its investigation of the recovered autonomous intrusion workspace.

  2. 2026-11-09

    Infosecurity Magazine reports the findings and qualifies the access and victim-count claims.

Technical breakdown

  • An exposed directory contained session transcripts, tool staging, credential caches, reconnaissance output and database dumps.
  • The operator supplied subnet goals and exclusions; the agent selected exploitation paths and executed consequential intrusion actions.
  • The reported agent adapted after endpoint defenses blocked a tunneling tool by writing and deploying a replacement.
  • The target-attribution list and patient-data access claims are not independently confirmed victim or exfiltration counts.

Authorization boundary

Where the authorization boundary should have been

This incident is categorized as Governance bypass. The relevant Permission Protocol gate is Tool-Call Gate. The read is conditional: the block only applies where the real action boundary is routed through a gate.

If enforced at
Agent-operator network/tool execution and victim identity/data boundaries
Still needs
PP is not perimeter defense. Patching, strong credentials, segmentation and endpoint controls remain necessary. It cannot constrain an attacker-controlled tool path outside its enforcement boundary.
Receipt required for
Authorized target-specific exploitation, credential use, bulk data retrieval and remote tool deployment in an integrated environment.

A victim's internal authority gate cannot constrain an external malicious agent harness that never routes its actions through that gate.

Start small

Put the relevant gate at this action boundary.

This incident maps to Tool-Call Gate. Start with the boundary that controls the actual action, then require a signed receipt before execution.

Replay this incident with a signer in the loop