The public record · Filed as detections land · Not legal advice

The AI-Vendor Terms-Change Feed

Every time a tracked AI vendor changes the clauses that touch your data — training-data use, retention windows, subprocessor swaps, license flips — it will be logged here in plain language: what changed, and why it matters for builders.

Currently monitoring 13 AI vendors, daily. No public detections have been filed yet — the first verified change will be published here with the vendor, clause type, and before/after summary. Below is exactly how each entry will read.

§1 Vendors under watch

OpenAIAnthropicGoogleMetaMistral AIStability AIMidjourneyRunwayKlingxAICoherePerplexityReplicate

Pages checked daily: terms of service, privacy policies, license pages, and usage policies where verified. Watch your own vendors →

§2 How entries will read

The three entries below are illustrative examples of the feed's format — they are not real detections. When TermsWatch files a real change, the entry will carry a detection date and a link to the vendor's actual page.

Illustrative example · Training-data use · Severity: HIGH

Example vendor updates its training-data clause

What changed: the terms would shift from "we do not use your content to train our models" to "we may use content you provide to improve our services unless you opt out." A single sentence changes the default for every user.

Why it matters for builders: your confidential prompts, uploads, and customer data could become training material unless you find and flip the opt-out — and it may not apply to every plan tier. This is the single most important clause for AI buyers.

Illustrative example · Data retention · Severity: MEDIUM

Example vendor extends its retention window

What changed: a 30-day retention window would become indefinite retention "for legitimate business purposes," with deletion-on-request gaining new exceptions.

Why it matters for builders: longer retention means your inputs live on the vendor's servers longer — a wider window for breaches, subpoenas, and internal reuse. Check whether the change applies to your plan before your next data push.

Illustrative example · Third-party sharing · Severity: MEDIUM

Example vendor swaps in new sub-processors

What changed: the sub-processor list would gain two new infrastructure providers, and "we don't sell your data" would gain carve-outs for service providers.

Why it matters for builders: your data now travels to more companies — each one a new breach surface and a new jurisdiction. If your privacy policy promised customers otherwise, you have a compliance exposure to fix.

§3 Get the alert before the entry

The public feed is the record. Free accounts watch 5 vendors with a weekly digest; Pro ($29/month) adds instant email + webhook alerts for every detected change, and Team ($79/month) adds API access and 5 seats. See pricing →

§4 Feed FAQ

What is the TermsWatch AI-vendor terms-change feed?

A public log of detected changes to the terms of service, privacy policies, and license pages of major AI vendors. Every entry names the vendor, the clause type, what changed, and why it matters for builders — in plain language, not legalese.

Which vendors are monitored?

Thirteen seeded AI vendors are monitored daily: OpenAI, Anthropic, Google, Meta, Mistral AI, Stability AI, Midjourney, Runway, Kling, xAI, Cohere, Perplexity, and Replicate. Users can add any other AI vendor by pasting its terms, privacy, or license page URL.

How often is the feed updated?

Vendor pages are checked daily. New public entries appear here as detections land; Pro users get instant email and webhook alerts for every detected change in their watchlist.

Is the feed free?

Yes — the public feed is free to read, no signup required. Watching your own vendors with alerts starts on the free plan (5 vendors, weekly digest), with Pro at $29/month and Team at $79/month.

Share this feed:

X · Facebook · LinkedIn · WhatsApp

This feed is informational, not legal advice. Detection dates and vendor names on real entries reflect what the platform observed, not legal conclusions.