Applied Ingenuity releases Project Chaperone to audit AI agent workflows
Start here
Applied Ingenuity has released Project Chaperone, an open-source harness that lets agentic AI call tools while enforcing high-level policies such as "no personal data may leave the network." A startup audit checks every route of all workflows for rule violations and asks a human to approve scoped exceptions, which are recorded per project; runtime enforcement then relies on data labels like UNTRUSTED and PERSONAL plus module properties such as OUTPUT and LEAVES_NETWORK. The repository ships five shared default rules and four examples, including deliveries, share_notes, ask_my_mail and safe_rag. An LLM is optional: a module may call one, and a state may let an LLM choose the next state in the workflow. Every setup except the demo requires users to supply their own LLM, mail or calendar, and the mail example reads the inbox over IMAP read-only, never changing, moving or sending a message.
Generated from reports · updated 10 min ago
Report timeline
Follow the reports to see the event from different sides.
- Hacker News · AIProject Chaperone: a rule-enforcing harness that audits AI agent workflows before they run
Applied Ingenuity released Project Chaperone, an open-source harness that lets agentic AI call tools while enforcing high-level policies such as no personal data leaving the network. A startup audit checks every route of all workflows for rule violations and asks a human to approve scoped exceptions, which are recorded per project; enforcement at runtime then relies on data labels like UNTRUSTED and PERSONAL plus module properties such as OUTPUT and LEAVES_NETWORK. The repo ships five shared default rules and four examples, including deliveries, share_notes, ask_my_mail and safe_rag, and an LLM is optional.
Heat for this event
Not enough continuous observations yet to draw a trend.