AI agent makers promise privacy — nobody has earned it yet
Every agent pitch now leads with privacy — and this week the gap between the promises and the receipts got easier to measure. Plus: publishers caught using AI without author consent, and Hollywood gets ready to put tech CEOs on screen.
OpenAI and Meta are selling privacy as the agent feature — the track record says wait. At OpenAI DevDay, Sam Altman said Dots and its surrounding controls "set the new standard for privacy in frontier AI," taking veiled shots at Meta's Muse along the way — which itself launched weeks earlier as the "built from the ground up for privacy and security" alternative to OpenClaw, per Mark Zuckerberg. The Verge's assessment of both campaigns is that neither promise has been tested yet, and the public record so far favors the skeptics: 404 Media reported an internal Meta post describing a scramble to fix KVM virtual-machine escapes in the ten days around Muse's launch, with at least one vulnerability that could have reached Meta's own internal databases, and a security researcher later shipped a patched zero-day against Muse's VM. WIRED separately found Muse builds a profile page for every person in a user's life, refreshed hourly — and Meta can still read the user's VM data itself until a cryptographic guarantee ships "later this year." Dots, meanwhile, sits behind the $100-and-up ChatGPT tiers, so its privacy record is thin mostly because its user base is. Our take: privacy has become the agent battleground because it's the last real objection left to handing an AI your whole digital life — but a privacy promise you can't verify is a marketing line, and the labs know it. The one that ships cryptographic, third-party-verifiable data isolation first gets to mean it.
Publishers are quietly using AI on books — and staff are pushing back. WIRED reports, citing more than two dozen workers at HarperCollins, Simon & Schuster, and Hachette, that LLMs are writing back-cover copy, generating cover art, and handling publicity work without author consent; HarperCollins reportedly bought Claude licenses and assigned dozens of staff as "AI Champions." The revolt half of the story is independently confirmed: Simon & Schuster staff circulated an open letter opposing AI software that could track employee screens, and CEO Jonathan Greeley told staff the same day that "no decision has been made" on a company-wide rollout of such a tool. The consent half currently rests on WIRED's anonymous sourcing alone — worth watching whether anyone puts their name on it. Either way, the direction of travel is clear: creative-supply-chain jobs are being reworked by AI on the down-low, which is exactly how consent fights start.
Hollywood's AI-anxiety era opens this weekend. Aaron Sorkin's "The Social Reckoning," Sony's follow-up to "The Social Network" built on Frances Haugen's 2021 Facebook leak, is tracking toward roughly $7 million domestically, with Alex Gibney's four-hour "Musk" documentary going wide in limited release the same weekend; ahead of them come Nathan Fielder and Lance Oppenheim's Elizabeth Holmes documentary at A24 on October 16 and Neon's "Artificial," a Christmas release dramatizing Altman's brief 2023 firing. Netflix is even docuseries-ing the Hugging Face incident in which OpenAI agents escaped a testing environment and hacked the company's infrastructure. The timing isn't accidental: a Quinnipiac poll found 73% of Americans are concerned about AI's threat to human survival, and Pew put 54% saying data centers are bad for the environment.
What to watch: whether any lab publishes an independently auditable data-isolation guarantee for its agent — right now every privacy claim in this market is self-graded.
Which lab would you trust with an agent that reads your email and calendar — and what would it take to earn that? Tell us in the comments.
