Meta's AI Terms Change October 30. ChatGPT Now Targets Ads From Your Prompts.

If your business still runs a Facebook or Instagram page after October 30, you will have agreed to something most people haven't read: Meta's right to scrape content tied to that account for AI training, with no toggle to turn it off.

What changed

Meta's updated commercial terms for business accounts take effect October 30. Nieman Lab reported that three legal experts read the language as letting Meta scrape content from any site "associated with" a business account on Facebook, Instagram, or WhatsApp — including news and client-facing sites — and feed it into AI training. After the story ran, Meta removed the literal word "train" from the terms, but the same experts say the underlying scraping permission didn't change, and one law professor flagged a separate clause that could let Meta register as a test user to get past a site's own login wall. There's no opt-out. Keep using Meta's business tools past the 30th, and you've agreed.

Two more vendors made data-access calls the same week, in opposite directions. 404 Media found that ChatGPT's ad program can now target Free and Go tier users by prompt content — OpenAI calls these "context hints" — plus uploaded email and phone-number matching lists, though inclusion lists need at least 25,000 matched users and turning off personalization doesn't stop ads tied to your current chat thread. Anthropic went the other way, giving six staff from the nonprofit evaluator METR company hardware and internal systems access, with no veto over what they publish beyond narrow redactions for customer data and model architecture.

What to do Monday

None of this requires a lawyer. The skill is treating "read-only" or "we don't train on your data" as a claim about this week's terms, not a permanent property of the tool — and checking on a schedule instead of once at sign-up. Pick the handful of AI-connected tools your team actually touches daily, and find out which ones changed their data terms in the last 90 days without telling you.

A marketing lead at a 15-person skincare brand manages the company's Instagram and Facebook pages herself, and the brand's blog — the one that pulls organic traffic from ingredient breakdowns — is linked to those accounts for ad purposes. She has nine days to decide whether that link is a problem. If it is, the fix isn't closing the pages; it's checking which of the brand's content actually touches a Meta business account, and writing down that she looked.

A different version, same week

An operations director at a 200-person logistics company has a team drafting internal routing memos in ChatGPT's free tier, because nobody has bought paid seats yet. The ad-targeting change matters less for what Meta can scrape and more for what his team pastes into prompts — client names, rate sheets, anything that could become a "context hint" an advertiser paid to reach. The fix isn't switching tools this week. It's writing down what's safe to paste into a free-tier AI tool and what isn't, before someone types a client's pricing into a prompt because no one ever said not to.

Three vendors made three different calls this week about what they owe you. The skill that applies to all three is writing the policy down before the next deadline lands, not after.