# Inth now has stricter privacy audits and actions

*Inth strengthens privacy audit verification, evidence handling and repair recommendations to help teams make better-supported decisions about what to act on.*

Engineering | 2026-09-10T14:44:13.721Z

Since launching Inth's privacy audits, we've been working on giving teams more context to decide what to fix\. On September 8th, we merged 34 PRs focused on verification, evidence, recommendations and the testing behind them\. Existing rollout controls still apply\.

## More context behind each finding

Audits now check the original allegation, supporting evidence and whether the cited privacy requirement appears relevant\. Verification also covers medium\- and low\-priority code defects\.

Valid citations from connected files and follow\-up searches stay attached through verification and reporting, with source and line checks to reject invalid references\. Uncertain findings remain visible without automatically becoming required fixes\. These are technical review checks, not legal advice\.

[See what a finding looks like](https://inth.com/docs/code-audit#what-a-finding-looks-like)\.

## Recommendations tied to the evidence

We've strengthened the checks that preserve confirmed repairs through grouping and report generation\. Plans that drop those repairs are rejected, and unverified observations cannot become primary remediation decisions\.

The goal is to keep the recommended change tied to the evidence you're reviewing\. Our [report review guide](https://inth.com/docs/code-audit#review-the-report) covers how to assess findings\.

## Broader testing and better recovery

We've added more than 100 privacy test cases, including vulnerable and fixed examples across consent, logging, deletion, queues, tenant isolation, caching and AI data handling\. The set includes realistic applications and executable browser\-consent scenarios across several frameworks and languages\. See our [coverage and limitations](https://inth.com/docs/code-audit#coverage-and-limitations)\.

Saved results can be replayed without another paid model run\. Failed runs and unknown costs remain visible\. We've also added bounded recovery for malformed responses, retained completed verifier output and improved timeout handling and cleanup\. Automated live evaluation checks remain opt\-in\.

False confirmations and recommendation\-quality issues remain; broader testing isn't a guarantee of accuracy\. For your own browser checks, see our [Playwright consent\-testing guide](https://inth.com/blog/test-consent-gates-nextjs-playwright)\.

## Turn findings into supported decisions

We want your team to spend more review time deciding what to change, with the evidence and repair instructions there to support you\. Keeping confirmed findings separate from open questions gives you a clearer basis for choosing what to act on and what to investigate further\.

[Open your audits](https://inth.com/dashboard) to review your findings and decide what to work on next\.

New to Inth? [Connect a repository](https://inth.com/docs/code-audit#connect-a-repository) and [run a full repository audit](https://inth.com/docs/code-audit#run-a-full-repository-audit)\.