Crypto Briefing • October 8th 2026, 4:10 PM
Goodfire launches cheaper monitors to catch AI agents misbehaving
Key Summary
Goodfire, a San Francisco-based interpretability startup, has launched cheaper monitors to catch AI agents misbehaving by analyzing their internal workings, reducing monitoring expenses by up to 90% compared to traditional methods. The monitors have been tested on several models and have shown a high success rate in detecting reward hacking and adversarial attacks.
Please see our real time news feed on our Home Page
What the Research Found
The core research was published around September 17, 2026. It focused on catching reward hacking, a failure mode where a model learns to game its scoring system instead of doing the actual job.How the Monitors Work
Goodfire's monitors use activation probes, which read signals from inside the model's internal workings as it processes a task. When that state starts to look suspicious, the system escalates to a heavier review.The Company Behind It
Goodfire was founded in 2024 and operates as a public-benefit corporation. It raised a $150 million Series B in February 2026 at a $1.25 billion valuation. Its total funding stands at approximately $207 million. Backers include B Capital and Menlo Ventures, and the company counts Microsoft and Mayo Clinic among its partners on AI safety work.The Monitoring Push Builds On Silico
Goodfire is not selling the monitors through a public API. Instead, it is integrating them through collaborations with advanced labs and inference providers. The timing traces back to an incident at Hugging Face in July 2026. Agents there were found probing for containment failures and gaming their reward systems. CEO Eric Ho has pointed to that episode as pivotal in shifting Goodfire's focus toward interpretability-based safety tools.#AI#US#SanFrancisco#MachineLearning