Guides · ← All guides
AI Vendor Terms Changes in 2026
Published: November 7, 2026 · Reviewed November 2026 · 8 min read
This guide is informational, not legal advice.
The short answer
AI vendor terms change constantly and quietly — training clauses, retention windows, and sub-processor lists flip with nothing more than a new "last updated" date. TermsWatch checks 13 tracked vendors daily and files every detected change in its public feed. As of November 2026 the feed is monitoring daily with no public detections filed yet; this page is the living log around it, starting from the documented flips TermsWatch's own guides record. Feed state as published on TermsWatch's feed page, November 2026.
What does a terms-change entry look like?
Every real entry in the feed names the vendor, the clause type, what changed, and why it matters for builders — in plain language, with a detection date and a link to the vendor's actual page. The feed's three published format examples (clearly labelled illustrative on the feed itself, not detections) show the shapes that matter most:
| Clause type | Severity | The change shape |
|---|---|---|
| Training-data use | HIGH | "We do not use your content to train our models" becomes "we may use content you provide to improve our services unless you opt out" |
| Data retention | MEDIUM | A 30-day retention window becomes indefinite retention "for legitimate business purposes," with new deletion exceptions |
| Third-party sharing | MEDIUM | The sub-processor list gains new infrastructure providers and "we don't sell your data" gains service-provider carve-outs |
Those are examples of the format. The dated entries below are different: each one is a change TermsWatch's published guides document with a real date, and each links to the guide that records it.
Which vendor terms changes are already on the record?
| Date | Vendor / event | What changed |
|---|---|---|
| March 2023 | Zoom | Terms updated with broad rights over customer content, including use for training and tuning AI models. Nobody noticed for nearly five months. |
| August 2023 | Zoom | Within a week of the changes surfacing publicly, Zoom revised its terms twice — first adding "without your consent" qualifiers, then stripping the AI-training permissions from customer-content terms entirely. |
| May 2024 | Slack | Privacy principles stated customer data — messages, content, files — is analyzed to develop AI and ML models. Opting out required a workspace owner to email Slack with the subject line "Slack Global model opt-out request." Slack later clarified customer data was not used to train generative AI products, only traditional ML features. |
| August 2024 | EU AI Act | The EU AI Act took effect — one of the regulatory shifts that keeps rippling into vendor terms documents. |
| January 2026 | California AB 2013 | California's AB 2013 training-data transparency requirements took effect, pushing training-data disclosure into vendor documentation. |
Sources on TermsWatch: the Zoom entries are documented in When AI Terms Change Silently, the Slack entry in Who Owns What You Type Into an AI Tool?, and the two regulatory dates in the silent-changes guide. No entry appears in this log without a published TermsWatch source behind it — and no detection will ever be backdated or invented to fill the table.
Why do the important changes arrive silently?
Because nothing forces a fanfare. Most terms allow the vendor to update policies with notice that amounts to a changed date line, and continued use counts as acceptance. The Zoom case is the proof: one of the most-read terms documents in the world carried new AI data rights for nearly half a year before public scrutiny found them. The defence is not reading harder once — it is watching continuously, with diffs. "The terms changed" is a useless notification; what changed, in which document, with the old and new text side by side, is an actionable one. That diff-first standard is what the TermsWatch feed is built to file, and Which AI Tools Train on Your Data? shows the document set being watched.
How do you get alerted before the next change lands on you?
Three levels, all starting free. Read the public feed — free, no signup, filed as detections land. Create a free account to watch 5 vendors with a weekly digest. Or go Pro ($29/month) for instant email and webhook alerts on every detected change in your watchlist, with Team ($79/month) adding API access and 5 seats — full details on pricing. Whichever level you pick, save a dated copy of every vendor's current terms today: without a baseline, you can never prove what changed or when.
Frequently asked questions
What is the AI vendor terms changelog?
A dated, living log of changes to AI vendors' terms of service, privacy policies, license pages, and usage policies — training-data clauses, retention windows, sub-processor swaps, and license flips — each entry naming the vendor, the clause type, and what changed in plain language.
How often do AI vendors change their terms?
Constantly. Vendors shipping new AI features every quarter rewrite their terms every quarter, and the changes that matter most are rarely announced — Zoom's March 2023 AI training language went unnoticed for nearly five months.
Has TermsWatch detected any public changes yet?
Not yet. The public feed monitors 13 vendors daily and states honestly that no public detections have been filed; the first verified change will be published there with a detection date and a link to the vendor's actual page.
What kinds of changes matter most?
Training-data use first — a single sentence can change whether your inputs train a vendor's models. Then retention windows, third-party sharing and sub-processor swaps, and license changes that alter what you may ship.
How can I watch my own vendors?
A free TermsWatch account watches 5 vendors with a weekly digest. Pro at $29/month adds instant email and webhook alerts for every detected change, and Team at $79/month adds API access and 5 seats.
Why not just re-read the terms every quarter?
A quarterly read misses the change that landed the day after, and without a saved baseline you cannot prove what the old text said. Continuous watching with diffs turns each change into a reviewable record instead of a silent date bump.
Read the public terms-change feed →
More from Ingenious AI: the fleet behind TermsWatch builds one focused tool per failure point — the full map is on the hub.