Release history
What’s New
Every change we ship to TermsWatchdog, with the date it went out. We include the fixes and the known problems, not just the new features — a transparency product that is not transparent about its own behaviour would be worth very little.
Last updated 31 July 2026 · currently on v2.1.1
Known issues
Things we know about and have not fixed yet. Removed from this list once resolved.
Reports published before 29 July 2026 use an earlier scoring method
Ratings issued before that date were produced by a different method and are not directly comparable to current ones. Re-analyse a tool to get a rating on the current scale. The analysis date is shown on every report.
Tools without publicly reachable policies are rated Unknown
If a vendor keeps its terms behind a login wall or at a non-standard URL we cannot read, the tool is marked Unknown rather than estimated. Unknown means undeterminable, not high risk. We would rather say we do not know than guess.
Belonging to more than one team can affect team features
If your account is a member of two or more teams, team-level features may not resolve correctly. Contact support and we will sort out the account.
Email security scanners can trigger unsubscribe links
Some corporate mail filters pre-open links in messages to check them, which can action an unsubscribe link without you clicking it. If you stop receiving change alerts unexpectedly, re-enable them from the report page.
Monitoring correctness
v2.1.1New features
- Individual tools can now be excluded from automated monitoring.
Fixes
- Daily monitoring was repeatedly re-checking the same tools instead of working through the full list. It now advances across the whole corpus.
- Policy documents that had not actually changed no longer register as changes, so change alerts reflect real edits.
More of every report, free
v2.1.0We moved the line between free and paid so that the ratings themselves are public.
New features
- All 15 risk categories now show their rating and confidence score without an account. Previously only three categories were visible.
- The written assessment for every category is now previewed rather than hidden.
- Verbatim quotes from the vendor’s own policy documents remain available to account holders — they are the evidence behind each rating.
- Reports can now be read by search engines and AI assistants, so a TermsWatchdog rating can be found and cited rather than sitting behind a gate.
Fixes
- Document upload showed an "Upgrade to Pro" prompt to Pro, Team and Enterprise subscribers who already had access. Upload was always permitted — only the prompt was wrong.
- The document uploader now lives on the Account page, with completed reports under Saved Reports.
Security update
v2.0.1Security
Resolved a set of issues identified during an internal security review, covering database access rules and the handling of submitted content. No action is required on your part, and no subscriber needs to change their password.
Ratings are now calculated, not judged
v2.0.0The largest change to how TermsWatchdog rates tools since launch.
Important
Ratings published before this release are not directly comparable to ratings published after it. Earlier ratings were a model judgement; ratings from here are calculated from the category findings by a fixed rule. Re-analyse a tool to place it on the current scale.
New features
- Consumer and enterprise terms are now rated separately. Enterprise ratings are derived only from enterprise documents — MSAs, DPAs, sub-processor lists and SLAs — and are never inferred from consumer terms.
- Where a vendor publishes no enterprise terms we can read, the report says so explicitly and tells you what to ask the vendor for.
- Every Moderate rating now names the specific finding holding it back.
Improvements
- The overall rating is calculated from the category findings rather than assessed separately. A serious finding in any of the four data-control categories now caps the overall rating regardless of how the others score.
- Ratings are now stable. Tools had been shifting between ratings between runs with no change from the vendor; that variability is gone.
- Re-analysis is skipped when a vendor’s documents have not changed, which makes monitoring faster and removes phantom "policy changed" alerts.
- A low rating now requires the vendor to have addressed a topic explicitly. Silence in a policy no longer earns credit.
Fixes
- The tools directory could show ratings that disagreed with the report it linked to.
- Data processing agreements published at non-standard paths, and terms hosted on third-party trust centres, were being missed during discovery.
Model refresh and view analytics
v1.13.0New features
- Reports on where transparency cards are being viewed geographically.
Improvements
- Moved to current-generation analysis models following the retirement of the previous ones.
Fixes
- Analysis no longer fails when a very long policy produces a truncated response.
Feedback, and a clearer free tier
v1.12.0New features
- Rate a report directly from the page and tell us where we got it wrong.
- When a search matches several products, you now pick from up to five candidates rather than accepting a guess.
- Reports show an estimated page count, so you can see how much fine print a rating is based on.
Improvements
- The free tier is now 5 reports in total rather than a monthly allowance, which matches how most people actually use it.
- Unknown ratings render in grey rather than red. Unknown means we could not read the policy, not that the tool is dangerous.
- Watchlist alerts now re-check stale reports before comparing, so alerts fire on genuine policy changes.
The public tools directory
v1.11.0New features
- A browsable directory of every published report at /tools.
Fixes
- Reports were not appearing reliably in search results. Addressed the underlying address and redirect problems.
- Tool logos were failing to load after a third-party logo service shut down.
Subscription repair
v1.10.0New features
- A confirmation email is now sent when a subscription is cancelled.
Fixes
- Repaired the subscription signup flow, including a payment webhook failure that could leave an account without the plan it had paid for.
- Signing up with an email address already in use now gives a clear message.
Industry Insights
v1.9.0New features
- A new articles section covering what AI terms of service actually say, with PDF downloads.
- Enterprise visitors can book a 20-minute demo directly.
Improvements
- Terms of Use expanded to cover subscription tiers and governing law.
One vendor, one report
v1.8.0New features
- Tools belonging to the same vendor are now grouped by parent company.
- Duplicate entries for the same product are merged, so a vendor has a single canonical report rather than several partial ones.
Improvements
- Faster, more reliable re-analysis when several tools are refreshed at once.
Personal Risk Profiles
v1.7.0Not every risk category matters equally to every organisation.
New features
- Choose which risk categories matter for your use case and get a personalised score alongside the overall rating.
- Five starter templates: AEC, Legal, Marketing, Healthcare and Procurement/IT.
- Free accounts keep a profile for the session; Pro saves one; Team and Enterprise support multiple named profiles.
- A dedicated enterprise page covering procurement use.
Fixes
- Donations failed on custom and some preset amounts.
Shareable report previews
v1.6.0Improvements
- Reports shared to LinkedIn and other networks now show a proper branded preview image instead of a bare link.
Security hardening
v1.5.0Improvements
- Vendors who publish their terms through third-party legal platforms are now analysed correctly rather than returning Unknown.
Security
A scheduled security review resulted in fixes across sign-in, email handling and request limits. No action is required on your part.
Smarter tool matching
v1.4.0Improvements
- Searching by product name now resolves to the right vendor domain before analysis. Searches like "Microsoft Copilot" reach the correct product rather than a near miss.
- Malformed or mistyped addresses are caught before they become a stored report.
Fixes
- Reports that could not be found now offer a new search rather than a dead end.
Policy dates and easier comparison
v1.3.0New features
- Reports now show the effective date of each policy document analysed, so you can see how current a vendor’s terms are.
- Comparison now works from your saved reports with a searchable picker rather than typed addresses.
- Cancel a subscription directly from the account page.
Improvements
- Longer analysis time limit, so vendors with very large policy sets complete reliably.
- All emails now include a plain-text version and a working unsubscribe header.
Fixes
- Team members on the free tier were not receiving the team’s level of access across all features.
Teams, watchlist and saved reports
v1.2.0The largest feature release since launch.
New features
- Team management: create a team, invite colleagues by email, and share reports across the team.
- Watchlist: track up to 25 tools and get an email when a rating changes.
- Saved Reports: keep snapshots of an analysis to compare how a vendor’s terms change over time.
- Side-by-side comparison of any two tools.
- PDF export for any report.
- Added a PII and sensitive-data risk category, and expanded compliance framework coverage.
Accounts and billing
v1.1.0New features
- User accounts, subscription plans and usage allowances.
- Branded email delivery for reports.
TermsWatchdog
v1.0.0The service launched under its current name, and settled into the shape it has today.
New features
- Full rebrand from the original working name.
Improvements
- Improved accuracy of the underlying policy analysis.
Policy discovery memory
v0.2.0New features
- Published Privacy Policy and Terms of Use.
Improvements
- Policy locations found once are remembered, so later analyses of the same vendor are faster and more reliable.
First release
v0.1.0New features
- Analyse any AI tool’s published terms and get a risk-rated transparency report.
- Per-category risk ratings with the policy text behind each one.
- Tools whose policies cannot be reached are marked Unknown rather than guessed at — a principle the product still holds to.
Spotted something we got wrong, or a problem not listed above? Tell us.