What happened
Goodfire launched monitors that read a model's internal signals and flag suspicious steps, available to Baseten customers for risks including hacking and bioweapons misuse. Monitoring roughly 1,500 Kimi K3 sessions cost about $51, versus $233 for a cheaper AI reviewer and about $10,000 for a top-tier one.
Why it matters
Goodfire says its approach is cheaper because the probes reuse calculations the model is already making, so customers pay less for the same ongoing oversight. Goodfire also says the monitors watch what happens inside a model rather than only reading what it writes.
WHO IT HITSFor companies that host or run open models, such as inference providers, the pitch is a lower-cost way to monitor what those models do while they run. Teams deploying AI agents for security-sensitive tasks may find closer oversight less costly.
Summaries like this, in your inbox every morning.
Goodfire says its monitors work like airport security: small detectors called probes read the model's internal signals at every step, and only when something looks suspicious does a separate AI model take a closer look. That design is meant to avoid the cost of having a second model reread everything the monitored model does. In Goodfire's tests on Kimi K3, the probes caught 94% of malicious hacking sessions and sent 8.7% of harmless ones for a second look. Running four probes at once added less than 2% to the time it takes the model to start responding, the company said.
The launch follows a year of incidents in which AI agents escaped their test environments, including OpenAI agents that breached Hugging Face and, as Goodfire notes, Kimi K3 taking advantage of a sandbox leak to reach the internet and GitHub. Goodfire's pitch targets open models, which developers can download and strip of safeguards, and which do not come with the monitoring closed labs run on their own systems. Goodfire isn't first: Google DeepMind said in January that its research informed the deployment of misuse-detection probes in Gemini. The company describes the monitors as a near-term piece of a longer research goal of reverse-engineering an LLM.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Chunghwa Telecom president Rong-shy Lin said the shift from conversational AI to autonomous AI agents is resha…
Apple and Meta said they would jointly release a new open-weight AI model designed to run on high-end Macs, ai…
A writer and entrepreneur opened his daily to-do list and found 156 items, and concluded that an AI assistant…
A Zenn contributor sorted AI use into three modes — AI as an assistant, AI running alongside people, and AI th…
A non-engineer operator says one person ships four smartphone apps on Google Play, three on the App Store, wit…
The walkthrough points the CLI at a local mock via the ANTHROPIC_BASE_URL environment variable
