Skip to content

Topic

AI training data

Text, image, audio and video content used to train machine learning models, and the sourcing, consent and licensing issues surrounding its collection.

Current stories

buildOne report1 publisher

Meta's October 30 Commercial Terms ask businesses on Facebook to let it crawl their sites for AI training

Meta's Commercial Terms take effect October 30 and ask Facebook business users to permit crawling of their websites for AI training. Meta says robots.txt controls stay as they are, but the permission sits in a contract.

Publishers:dev.to

Reality

Evidence55
Adoption
Not scored
Hype gap+10
Incentives
Not scored
Confidence50
leadershipOne report1 publisher

AI data offers for company Slack and Docs come with six-figure sums and 48-hour deadlines

AI labs and their data brokers are offering six-figure sums for live workplace data such as Slack DMs and Docs, Business Insider reported. Gartner projects AI data spending at almost $6.5 billion in 2027, so a company needs a policy on selling its data before the offer arrives.

Reality

Evidence50
Adoption55
Hype gap+20
Incentives70
Confidence55
productOne report1 publisher

Closing a startup now includes deciding whether to sell its Slack history to an AI lab

SimpleClosure says 47% of its startup wind-down deals now include a sale of code or workspace data. One founder netted about $25,000 for his code base, and a lawyer who advises AI startups calls such data "legally radioactive."

Reality

Evidence50
Adoption45
Hype gap+20
Incentives70
Confidence50
investConfirmed4 publishers

Lawmakers press Google to strip employee records from its $10 million Spirit Airlines data purchase

More than 100 lawmakers led by Sen. Elizabeth Warren asked Google and Spirit Airlines to keep former employees' records out of a $10 million data sale. So far the objection has cost Google one thing: an independent scrub of the data, a step it says it offered from the start. A bankruptcy judge takes up the sale at an Oct. 14 hearing.

Perspective Coverage

4 publishers
Builder
Builder 26%
Operator
Operator 45%
Investor
Investor 29%

Reality

Evidence62
Adoption
Not scored
Hype gap+15
Incentives60
Confidence58
buildConfirmed5 publishers

USA TODAY Co. asks a federal court to make OpenAI pay for training on 19 publications

USA TODAY Co. sued OpenAI for more than $250 million and an injunction over training ChatGPT on work from 19 publications. The case tests whether a publisher can force payment and limits on the training data that teams building on OpenAI's models depend on.

Perspective Coverage

5 publishers
Builder
Builder 28%
Operator
Operator 38%
Investor
Investor 34%

Reality

Evidence68
Adoption
Not scored
Hype gap+5
Incentives62
Confidence66
productOne report1 publisher

Meta says its Muse agent may still remember what it learned from deleted chats

Meta says its Muse agent "may still remember information it learned from what you deleted," and PCWorld reports that Muse data trains Meta's AI by default. Anyone connecting an inbox to it is making a choice that deleting a chat will not fully reverse.

Publishers:pcworld.com

Reality

Evidence45
Adoption
Not scored
Hype gap
Not scored
Incentives
Not scored
Confidence40
buildOne report1 publisher

Unsealed NYT v. OpenAI brief quotes Microsoft and OpenAI insiders on scraping and paywalls

New York Times lawyers quote a 2023 Microsoft memo calling AI scraping 'the largest theft of labor in human history' in briefs a court unsealed September 17. For teams training on unlicensed data, the case puts dataset custody and internal email in the same record.

Publishers:dev.to

Reality

Evidence45
Adoption
Not scored
Hype gap+15
Incentives70
Confidence40
buildOne report1 publisher

David AI's benchmark finds speech systems 25 points apart on Indic conversation

David AI's 21-language benchmark found speech systems' error rates spread an average of 25.4 points in five Indic languages, against 3.7 in the tightest five. In those languages, choosing a system is a 25-point decision that an English score can make look like a close call.

Publishers:runtimewire.com

Reality

Evidence45
Adoption
Not scored
Hype gap+10
Incentives60
Confidence50
productConfirmed8 publishers

Twitch's AI opt-out shipped after the training: consent as a retrofit

The toggle that stops Amazon training generative models on your streams is off by default, and Twitch's product chief says that is deliberate: if it were opt-in, nobody would opt in.

Perspective Coverage

8 publishers
Builder
Builder 24%
Operator
Operator 56%
Investor
Investor 20%

Reality

Evidence72
Adoption
Not scored
Hype gap+15
Incentives78
Confidence70
productConfirmed3 publishers

Used bookshops sell out by the pallet: provenance is now a line item, not a research problem

A single order matched a week of sales at one Northumberland shop. The legal machinery behind it turns clean pre-LLM text into a purchasable commodity with a jurisdictional price.

Perspective Coverage

3 publishers
Builder
Builder 30%
Operator
Operator 43%
Investor
Investor 27%

Reality

Evidence55
Adoption30
Hype gap+35
Incentives55
Confidence50
productConfirmed4 publishers

Google paid $10M for a dead airline's paperwork, and the flight attendants object

The Spirit Airlines bankruptcy auction put a price on a defunct company's internal records. The union's objection shows the scrubbing standard was written for customers, not staff.

Perspective Coverage

4 publishers
Builder
Builder 30%
Operator
Operator 39%
Investor
Investor 31%

Reality

Evidence72
Adoption35
Hype gap+15
Incentives70
Confidence65
leadershipOne report1 publisher

Oxford's Bodleian scans fed OpenAI's training set, internal documents show

Oxford let OpenAI train on Bodleian texts it digitised, with 125,000 dissertation images shared by June 2025, internal documents show. Oxford says staff never hid the training use, though its 2025 announcement spoke only of wider access.

Publishers:theguardian.com

Reality

Evidence58
Adoption22
Hype gap+8
Incentives62
Confidence60
productConfirmed7 publishers

Seattle Times and Newsday sue OpenAI and Microsoft over copyright, paywall bypass and lost revenue

The two papers say their paywalls were bypassed methodically, which puts the terms of access at issue in the US and touches a commitment OpenAI already signed in Europe. Microsoft went after the damages theories the same day.

Perspective Coverage

7 publishers
Builder
Builder 29%
Operator
Operator 42%
Investor
Investor 29%

Reality

Evidence70
Adoption
Not scored
Hype gap+10
Incentives55
Confidence65

Earlier coverage

  1. DOJ's 20-page brief backs OpenAI, argues AI training on copyrighted work is fair use

    Invest · September 2, 2026 · Confirmed4 publishers

  2. Four passing runs out of 80 separate first from second on Specific's private-code benchmark

    Build · September 12, 2026 · Confirmed3 publishers

  3. Microsoft measured a 93 percent click-through drop to nytimes.com from its own answer engine

    Product · September 17, 2026 · Confirmed6 publishers

  4. OpenAI's own executives call ChatGPT substitutive of the journalism it trained on

    Build · September 17, 2026 · Confirmed5 publishers

  5. Nadella testified that chatbots have "substituted" for publisher sites

    Security · September 17, 2026 · Confirmed3 publishers

  6. Snorkel sells about a tenth of itself for $350 million at a $3.5 billion valuation

    Invest · September 22, 2026 · Confirmed3 publishers

  7. Snorkel books a $375m run rate a year after switching from tooling to finished datasets

    Product · September 22, 2026 · Confirmed2 publishers

  8. Google bid $10 million for 600 million messages sent by Spirit Airlines employees

    Product · September 17, 2026 · Confirmed2 publishers

  9. Anthropic's claims portal puts the burden of proving reverted rights on the author

    Product · September 6, 2026 · Confirmed2 publishers