Governance Lag Monitor is being built to track what happens after frontier AI evidence is published. For each dated public source, it looks for a related control or decision and records the time between them.
Anthropic activated its AI Safety Level 3 Deployment and Security Standards for Claude Opus 4 even though it had not determined that the model definitively passed the ASL-3 capability threshold. It deployed the model under ASL-3 as a precaution because ASL-3 risks could not be ruled out.
What happened and when.
The public source, such as a system card, framework update, Model Report, AI Safety Institute publication, or regulatory filing.
A documented control, release decision, escalation, or the absence of a public response.
Time from evidence to response.
Whether the item is closed with a control, open, or being watched.
Citations to the supporting public record.
The preview shows the records being tracked now. Open items continue to accrue time until a public response is documented.
The source material is public and free to access. The product organizes it into searchable, analyst-reviewed records that teams can compare and export.
The app will connect findings to later controls and public commitments to later policy changes. It will also track regulatory deadlines and flag items with no documented response.
Searchable records, entity views, source links, and open-item status.
A monthly note on material changes, newly opened or closed items, and analyst updates.
A PDF briefing with an institutional source packet.
In development
The public preview includes selected entries and the method. Planned subscription tiers add the full tracker, saved views, alerts, exports, briefings, and integrations.
Selected records, methodology, and quarterly public notes.
The full tracker for individual users who monitor frontier AI governance.
Shared watchlists, source packets, and board-ready briefings for review teams.
Custom taxonomies and API access, with optional Syntony advisory support.
The Monitor uses public evidence only. Each record cites its source, including system cards, published frameworks, Model Reports, AI Safety Institute publications, and regulatory filings. Client-engagement information is never included.
System cards, frameworks, regulatory filings, and institute reports use different formats. The Monitor normalizes dates, entities, status, and sources so a reviewer can compare them.
The Monitor is designed for risk and procurement teams, insurers, and boards that need a current, sourced view of frontier AI evaluation and the governance responses that follow.