Across the seven cases below that carry a primary document, no measured loss of customers is documented in the sources reviewed, as of November 17, 2026, while a regulator or the company itself acknowledged concern in several.
This page covers the licensor's reputation only: the cases, the drivers, the mitigations and the case each side makes.
It does not claim that a license damages a brand, and not licensing is a valid outcome for any holder weighing AI training data licensing. Privacy of the people in the data, creator pay and the staff view have their own pages.
DataPacts Editorial Team dated and checked every statement on this page against the company's own document or a regulator's, and the page on how we source, date and check every figure sets out the method.
Documented Backlash Cases and What Each Company Said
Seven cases carry a primary document in which the company, a regulator or a representative body states what was changed and why people objected. The table lists them by date, as of 2026-11-17; the Twitch row concerns a platform training on its own users' data, not a sale.
| Date | Company | What the company or regulator documents | Stated by | Reaction on record | Source |
|---|---|---|---|---|---|
| 2023-08-07 (blog); 2023-08-11 (terms) | Zoom | blog on terms and AI features, updated 2024-02-07; Terms s.10.2 states no training of Zoom or third-party AI models on communications-like Customer Content; Update Notes: update to "clarify" data usage and "narrow the scope of Zoom's licenses" | Zoom | none documented in Zoom's own documents | zoom.com |
| 2024-02-27 | Automattic (WordPress.com, Tumblr) | public posts shared with select AI companies from sites not opted out; removal requested after new opt-outs | Automattic | not documented; recipients and value not disclosed | automattic.com |
| 2024-03-14 | Reddit, Inc. | FTC letter on a non-public inquiry into licensing user content to train AI | Reddit (S-1/A) | outcome not found | sec.gov |
| 2024-05-06 | Stack Overflow | OpenAI release: OverflowAPI access, attribution in ChatGPT; no value disclosed | Stack Overflow | not documented in primary sources | stackoverflow.co |
| 2024-06-06; 2024-06-10 | Adobe | post on the Terms of Use update: Firefly not trained on customer content; update note of 2024-06-10: "We've heard your concerns about our recent Terms of Use re-acceptance rollout" | Adobe | concerns acknowledged by Adobe | blog.adobe.com |
| 2024-11-19 | HarperCollins | per-title offer split 50-50, per-author opt-in addendum; licensee unnamed | Authors Guild (publishes the terms) | Guild views on the split, as the parent page records | authorsguild.org |
| 2026-08-20 | Twitch (Amazon; platform's own users, not a sale) | notice that the "Training for Generative AI" setting is on by default; advice to switch it off | Autoriteit Persoonsgegevens | advice to users | autoriteitpersoonsgegevens.nl |
Adobe's post of June 6, 2024 says the focus of the update was to be clearer about improvements to its moderation processes, that Adobe does not train Firefly Gen AI models on customer content, and that Adobe will never assume ownership of a customer's work.
A note dated June 10, 2024 says the company had heard concerns about the re-acceptance rollout.
Zoom's Terms of Service s.10.2, on a page effective August 11, 2023, state that Zoom does not use communications-like Customer Content to train Zoom or third-party AI models, and the Update Notes describe the change as a clarification that narrows Zoom's licenses.
No Zoom or Adobe document measures what users did. The program terms and record of the buyer are in Adobe's buyer profile.
Reddit disclosed the FTC letter of March 14, 2024 in its S-1/A. No outcome is on the record, and no user-reaction figure is.
None of the sources reviewed documents a loss of customers, a boycott or a share-price effect from a data license. This page relies on primary documents, and those do not measure a reaction.
The parent page lists the other objections, inquiries and rulings on record, among them the Meta complaints and the Spirit Airlines objections, which concern training on a platform's own users or a court-supervised sale and are not repeated here.
The cases above differ in kind: Zoom and Adobe changed terms for their own customers, Automattic and Reddit licensed user content, and HarperCollins and Stack Overflow concern content that creators or contributors wrote.
Stack Overflow's release of May 6, 2024 with OpenAI states that OpenAI will provide attribution to the Stack Overflow community within ChatGPT and gives no value. The deal terms are on the page about Stack Overflow AI licensing deals.
What Drove the Reaction: Changed Terms, Defaults and Silence
In the cases with a primary document, the trigger was how the change reached people, a terms update, a default that was already on, or a use nobody announced, rather than the license itself. The table sets four drivers beside the documents that show them, as of 2026-11-17.
| Driver | Case | What the document shows | Source |
|---|---|---|---|
| Changed terms | Adobe 2024-06; FTC staff 2024-02-13 | re-acceptance rollout, concerns acknowledged; staff: quiet permissive changes may be unfair or deceptive | blog.adobe.com; ftc.gov |
| Default on | Figma 2024-08-15; Twitch 2026-08-20 | training on by default for Starter and Professional plans; setting on by default, advice to switch off | help.figma.com; autoriteitpersoonsgegevens.nl |
| No way to say no | Automattic 2024-02-27; HarperCollins 2024-11-19 | opt-out added; per-author opt-in addendum | automattic.com; authorsguild.org |
| Scope not announced | Reddit 2024-03-14 | FTC non-public inquiry into licensing user content | sec.gov |
The Figma row concerns training on a company's own customers' content, and the Twitch row a platform and its own users, so both show the default-on driver without involving a sale.
The Adobe record shows a terms rollout that drew stated concern, and the Automattic and HarperCollins rows show the remedy of choice added to a license, either as a site-level opt-out or a per-author opt-in.
Reddit's letter shows a regulator asking about scope, and it is a question about a use of user content, not a finding. In every row the company or regulator speaks first, and the reaction column of the cases table stays empty where no primary document records one.
The FTC Office of Technology staff post of February 13, 2024 says it may be unfair or deceptive for a company to adopt more permissive data practices quietly.
It is a staff blog, not a Commission rule. The rules on changed privacy policies and model deletion are set out in FTC rules on AI training data.
What Companies Did About It: Disclosure, Opt-outs and Revenue Sharing
Companies on record answered with four tools, each with a stated limit: saying plainly what is and is not used, letting people opt out or opt in, naming attribution, and paying a share. The table sets each tool beside an example and its limit, as of 2026-11-17.
| Tool | Example on record | Stated limit | Source |
|---|---|---|---|
| Say plainly what is not used | Zoom s.10.2; Adobe post 2024-06-06 | covers the company's own models and systems, not third parties that collected the content elsewhere | zoom.com; blog.adobe.com |
| Opt-in only | Slack: trains generative AI on customer data only with affirmative opt-in consent | a statement by the company, not an audit | slack.com |
| Site-level opt-out | Automattic | removal is requested, no recipient has stated that it honors it | automattic.com |
| Attribution | Stack Overflow and OpenAI | how it is shown and enforced is not stated | stackoverflow.co |
| Share of revenue | Shutterstock Contributor Fund, 2022-10-25; HarperCollins split | does not answer an objection about consent; YouTube states it is not facilitating payments between third-party companies and creators at this time | shutterstock.com; authorsguild.org; support.google.com |
Automattic's post of February 27, 2024 says public posts go to select AI companies only from sites that have not opted out, and that the AI companies were updated about new opt-outs and asked to remove content from past sources.
The recipients are not named and no recipient has said that it honors the removal request. The opt-out and removal terms are set out in full on the page about Automattic (Tumblr, WordPress.com) data licensing.
Revenue sharing exists on both sides of the record. Shutterstock announced its Contributor Fund on October 25, 2022 and states that first earnings from AI dataset deals were issued to contributors in Q4 2022, checked 2026-11-09. The Authors Guild published the HarperCollins offer of November 19, 2024 as USD 5,000 per title, split 50-50 between publisher and each author who opts in.
YouTube states that it is not facilitating payments between third-party companies and creators at this time, checked 2026-11-09.
Each tool has a limit:
- An opt-out leaves content that others collected before or elsewhere.
- A statement of non-use covers only what the company itself controls.
- Payment does not answer an objection about consent.
A company that cannot recall a trained model also carries a trust issue, because one buyer program, Labelbox, states that models already trained on the data cannot be unlearned, checked 2026-11-17. The method for removal from datasets sits elsewhere: see platform settings for opting out of AI training.
Is the Reputational Risk Real or Overstated? The Case Each Side Makes
Both sides rest on the same record: one points to regulators and users reacting to how changes were made, the other to the lack of any documented loss and to licenses that continued.
The risk is real
- FTC staff said on February 13, 2024 that quiet changes to data practices may be unfair or deceptive.
- Adobe wrote on June 10, 2024 that it had heard concerns about its Terms of Use re-acceptance rollout.
- A regulator, the Dutch authority, told users on August 20, 2026 to switch a default setting off.
- Reddit disclosed in its S-1/A an FTC letter of March 14, 2024 on a non-public inquiry into its licensing of user content.
The risk is overstated
- No loss from a data license is documented in the sources reviewed.
- Reddit's 10-Q for the quarter ended September 30, 2026, filed October 29, 2026, reports agreements with its two main licensing partners that are approaching the end of their current terms, and Other revenue of USD 47,961 thousand, a line that includes products sold to users.
- Zoom and Slack answer the training question with a plain no or an opt-in in their current terms, checked 2026-11-09, and no document records a loss after they did.
- A payment route has existed since the Shutterstock fund of 2022.
Neither side shows a measured effect: the real-risk list rests on stated concern and regulator attention, and the overstated list rests on absence of evidence, which is weaker than evidence of absence.
The disagreement is whether documented reactions to terms changes predict reactions to licenses made with notice. The staff view sits on its own page, and the contract side of the question is set out in customer consent and confidentiality clauses before licensing data for AI.
Licensing and Trust: Benefits, Risks, Downsides and Alternatives
A license can bring income and a defined set of limits, and it can also draw objections that no contract term answers. The lists below set the two against each other, as of November 17, 2026.
Benefits
- Income and use limits written down, as in Reddit's disclosed agreements and the HarperCollins split.
- A license announced with a stated opt-out or opt-in gives users a defined choice.
Risks
- Objections from users or a regulator when terms change or a default is on.
- An inquiry, as in the FTC letter to Reddit.
- Users or creators reading attribution or pay as inadequate, which no attribution clause or fixed split settles in advance.
Downsides
- No recall from a model after training is a trust issue as well as a privacy one.
- Income that may end, as Reddit's 10-Q says of its two main agreements.
Alternatives
- License only data whose people agreed or were told.
- Opt-in per person.
- Collective licensing.
- Not licensing at all, which keeps every other option open.
The full comparison is on the page about alternatives to selling data, including not selling at all.
How Reputation Fits the Ethics of Selling Data
Reputation is one of four ethics questions about selling data for AI training, next to privacy of the people in the data, the staff view and fair treatment of creators. The parent overview covers ethics and risk of selling data for AI training in full.
- Privacy: the people in the data
- Staff: staff inside the archive
- Creators: creators whose works are in the data
- Deals and platforms: the Reddit filings in full; Zoom's terms and recordings in detail; how platforms license user content
- Court-supervised sale: Spirit Airlines' data archive sale
- Decision: the full decision
What This Page Does Not Cover
This page covers the reputation of the company that licenses data, not the buyer's licensing risk, ownership disputes, or brand risk from AI outputs. The Spirit Airlines objections stay on the parent page and the deal page. The page is information, not legal advice.
Frequently Asked Questions
Has licensing data to AI cost companies customers?
No loss of customers from a data license is documented in the sources reviewed for this page, as of 2026-11-17. Reddit's 10-Q for the quarter ended September 30, 2026 reports Other revenue of USD 47,961 thousand, but the line includes products sold to users and does not isolate licensing.
What do users object to most when a company licenses data to AI?
In the cases with a primary document, users and regulators objected to changed terms and default-on settings more than to the existence of a license. Adobe acknowledged concerns about its re-acceptance rollout in June 2024, FTC staff warned about quiet changes in February 2024, and the Dutch authority advised Twitch users in August 2026 to switch off a default setting.
How do companies limit the reputational risk of a data license?
On record they state plainly what is not used, offer an opt-out or opt-in, add attribution, or share revenue, and each has a stated limit. Zoom's terms exclude training on communications-like content, Automattic shares posts only from sites that have not opted out, Slack requires opt-in consent, and Shutterstock announced a contributor fund in October 2022.