Uber CISO Philip Martin tells Anthropic frontier models need insider-threat-style monitoring
Uber CISO Philip Martin told Anthropic that frontier models need real-time monitoring controls like insider-threat programs, and said Uber pairs an LLM gateway with open-source ADR tooling.
Why it matters
No token, market or regulatory consequence is on offer here. The relevance is operational: an enterprise security chief at a large company is arguing that model capability has outrun the controls used to supervise it, and Uber describes deploying controls in production. It is a single sourced interview with no independent confirmation.
Key facts
Uber CISO Philip Martin said frontier AI systems need monitoring controls comparable to insider-threat programs companies run for employees.
Reported by Blockchain.News
Martin said increasingly capable models can be directed to act outside expected constraints, making real-time detection and intervention a requirement for safe deployment.
Reported by Blockchain.News
The interview was published by Anthropic on Oct. 13, 2026.
Reported by Blockchain.News
Martin said Uber applies the capability to security reviews and threat modelling by feeding models code, design documents, Jira issues and historical service context.
Reported by Blockchain.News
Martin described a future in which each service has an agent that exchanges threat-model information with other agents, mirroring how production services interact.
Reported by Blockchain.News
The interview says Uber worked with Anthropic on Enterprise Frontier Safeguards, which let Uber keep control of its data and decide how to respond when a safeguard is triggered.
Reported by Blockchain.News
Martin said Uber's AI security setup combines its LLM gateway, which sets a perimeter for AI use, with open-source AI Detection and Response tooling and Anthropic's safeguards.
Reported by Blockchain.News
Uber's ADR repository states the tool is deployed in production at Uber and is designed to discover AI tools and observe their use.
Reported by Blockchain.News
What happened
Uber chief information security officer Philip Martin said in an interview published by Anthropic on Oct. 13, 2026, that frontier AI systems need monitoring and intervention controls of the kind companies already run to catch risky employee behaviour. He said increasingly capable models can be steered to act outside expected constraints, which makes real-time detection a requirement for deployment rather than a later check. At Uber, Martin said that capability is already feeding security reviews and threat modelling, with models given code, design documents, Jira issues and historical service context, and he described a future in which every service carries an agent that exchanges threat-model information with other agents the way production services interact. The interview also says Uber worked with Anthropic on Enterprise Frontier Safeguards, which let Uber retain control of its data and choose how it responds when a safeguard fires. Martin said Uber's setup combines its own LLM gateway, which sets a perimeter for AI use, with its open-source AI Detection and Response tooling and Anthropic's safeguards; Uber's ADR repository separately states the tool runs in production at Uber and is built to discover AI tools and observe their use. Blockchain.News is the only outlet with this story, so every claim here rests on that single report.
How the story developed
- First report
First report by Blockchain.News
Uber CISO says frontier AI needs insider-threat monitoring
- Update
Anthropic publishes interview with Uber CISO Philip Martin
Martin argues frontier models need insider-threat-style monitoring and describes how Uber combines an LLM gateway, open-source ADR tooling and Anthropic safeguards.
Affected entities
Original sources
| Publisher | Report | Role | Published |
|---|---|---|---|
| Blockchain.News Crypto media | Uber CISO says frontier AI needs insider-threat monitoring | Original |