Dobby is a data policy platform designed to provide evidence for the decisions made by AI systems, particularly where those decisions are difficult to prove or audit. The platform addresses the challenge organizations face in demonstrating compliance with data policies when AI agents make decisions about data access, movement, and oversight—actions that are often opaque to human reviewers. Dobby produces audit-ready evidence packages that procurement teams and auditors can accept, aiming to close the gap between AI activity and regulatory or organizational requirements.
The platform operates out-of-band, connecting to AI agents, workflows, and pipelines without sitting in the request path. It reads run telemetry after the fact, ensuring that it does not interfere with operational workflows. Dobby scans every run against a four-layer policy hierarchy—platform, organization, tenant, and process—with the strictest rule prevailing. Each run is checked against activated compliance frameworks using deterministic rules and AI evaluation, resulting in a four-state verdict: compliant, violated, needs review, or unverifiable. The platform then exports a self-contained evidence pack per framework, including a control matrix, gap report, findings, and a signed manifest, packaged for procurement and audit review in formats such as PDF and ZIP.
Dobby is framework-agnostic and allows each tenant to activate the compliance modules relevant to their domain. The current flagship module, Fintech AI Evidence, supports compliance with the EU AI Act, DORA, and SOC 2 standards, targeting AI vendors selling into banks and fintechs as well as risk teams conducting vendor reviews. The platform is non-custodial by design; for regulated modules, raw payloads are enforced into a customer-owned store, with this routing available as an opt-in feature elsewhere. Additional modules for broader enterprise data policy enforcement are mentioned as forthcoming, with anticipated support for ISO 27001, NIST CSF, and GDPR frameworks.
Dobby positions itself as a solution for organizations that need to demonstrate AI compliance and governance, particularly in regulated industries where audit trails and evidence packs are required for procurement and review processes.
dobby-collector is a LLM eval & observability product. It focuses on monitoring and governing the behavior of AI agents by collecting and streaming telemetry data. It is built as an open-source project for AI developers and researchers. dobby-collector is open source under the MIT license. dobby-collector is available on the web, the command line, and API, and it can be self-hosted.
It is developed by gil-dobby, and the product first shipped in 2026. Key capabilities include telemetry streaming, agent observability, and LLM call tracking.
Latest indexed changes and source events
Other apps tracked under the same category.
