---
id: "2026-10-08-goodfire-activation-probe-monitors"
url: "https://postcutoff.com/e/2026-10-08-goodfire-activation-probe-monitors/"
as_of: "2026-10-08T23:45:00+02:00"
date: "2026-10-08"
date_precision: day
category: policy-safety
importance: 3
confidence: medium
status: [Partly confirmed]
sources: 1
editor: Adam Bicz
human_review: null
version: "2026-10-08"
---

As of: 2026-10-08 23:45 CEST.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code). Human editor: Adam Bicz.
Canonical page: https://postcutoff.com/e/2026-10-08-goodfire-activation-probe-monitors/

# Goodfire launches 'inside-out' activation-probe monitors for AI agents

Full title: Goodfire launches 'inside-out' activation-probe monitors for AI agents: 94% of malicious hacking sessions caught at ~$51 vs ~$10,000 for a frontier-model monitor

On Oct 8, 2026 interpretability startup Goodfire released monitors that read a model's internal activations with small classifiers ("probes") instead of having a second AI reread every output. In Goodfire's test on Kimi K3, about 1,500 agent sessions, the probes caught 94% of malicious hacking sessions and flagged 8.7% of benign ones. Monitoring cost about $51, against about $233 for a cheap monitor model and about $10,000 for a top-tier one. The monitors are offered to Baseten customers.

## Key facts
- Method: probes on intermediate activations the model already computes; CEO Eric Ho: 'All you're doing is taking the intermediate neural activations that it's already computed and then running a classifier over these internal computations'
- Kimi K3 test (~1,500 sessions): 94% of malicious hacking sessions detected; 8.7% of harmless sessions flagged for second review; <2% added latency with four probes running at once
- Cost for the test set: ~$51 with probes vs ~$233 with a cheaper monitor model vs ~$10,000 with a top-tier monitoring model
- Also tested on GLM-5.2 and in a Google DeepMind Gemini deployment (TechCrunch)
- Configurable categories: offensive hacking, chemical/biological weapons misuse, reward hacking
- Availability: to customers of Baseten, which hosts open-weight models (probes need access to activations, so they suit self-hosted or open-weight models)
- All figures are Goodfire's own; no independent evaluation yet

## What happened

Goodfire, the interpretability startup co-founded by CEO Eric Ho and CTO Dan Balsam, turned activation probes into a commercial
monitoring product for agent deployments. It is offered through the model host Baseten. It targets the misuse categories that
labs worry about most for open-weight models: cyber offense, CBRN and reward hacking.

Confidence is medium: only TechCrunch's report was read, and Goodfire's own write-up was not located.

## Why it matters

Open-weight models with frontier-level cyber skills (GLM-5.3, Kimi K3) cannot be monitored at the API like closed models. Monitoring
through the model's internals is about 200x cheaper than a frontier-model monitor, which could make safety monitoring of self-hosted
agents affordable. It is also one of the first commercial uses of interpretability.

## Your AI and this story

- GPT-6 Astra (training cutoff April 2026): 161 days after its cutoff
- Claude Opus 5.5 (training cutoff June 2026): 100 days after its cutoff
- Gemini 3.8 Flash (training cutoff March 2026): 191 days after its cutoff
- Grok 4.7 (training cutoff May 2026): 130 days after its cutoff

## Sources

1. [TechCrunch: Goodfire says its new 'inside-out' monitors catch rogue AI agents at a fraction of the cost](https://techcrunch.com/2026/10/08/goodfire-says-its-new-inside-out-monitors-catch-rogue-ai-agents-at-a-fraction-of-the-cost/) (techcrunch.com, press)

## Changes

- 2026-10-08 (filed): Created

## Related

- 2026-09-29: [Anthropic: open-weights GLM-5.3 nearly matches Mythos Preview at exploit development](https://postcutoff.com/e/2026-09-29-anthropic-glm-5-3-spread-of-cyber-capabilities/index.md)
- People: [Eric Ho](https://postcutoff.com/person/eric-ho/)