Daily AI news, sourcedWhere it stands
As of Oct 9, Goodfire launched activation probes that watch an AI model’s internal signals while an agent runs, escalating to a separate model only when a probe flags risk, and made them available to Baseten customers, per TechCrunch.
- Oct 9Reported by TechCrunch
Goodfire ships inside-model probes that monitor AI agents cheaper than second-model watchers, per TechCrunch
Goodfire launched activation probes that watch an AI model’s internal signals while an agent runs, escalating to a separate model only when a probe flags risk, and made them available to Baseten customers, per TechCrunch. Baseten users can select risks such as offensive hacking, chemical/biological misuse, and reward hacking, plus responses from logging to human review or refusal. On Kimi K3 tests, monitoring about 1,500 sessions cost roughly $51 versus $233 for a cheaper external checker and about $10,000 for a top-tier one, with probes catching 94% of malicious hacking sessions and escalating 8.7% of harmless ones, while four probes added under 2% time to first token, Goodfire said.
Source: TechCrunch
- Official
- An institution announces something that has already happened, through its own channel or a founder or executive account.
- Statement by …
- A named person's views, predictions, plans or comments, including remarks reported by media when the original is unavailable.
- Reported by …
- Only media or another third party is saying it. We name the outlet.
- Unconfirmed
- Circulating online (posts, comments, leaks). No company, named person or outlet has confirmed it.
Items older than 72 hours never go in Top stories and show their original date.