{"slug": "always-on-always-watching", "title": "Always On. Always Watching.", "summary": "Meta disabled its AI glasses' camera if the recording LED is tampered with, but the same week it tested an always-on 'super-sensing' mode that would keep the camera running continuously, and CEO Mark Zuckerberg questioned whether the LED needs to stay on. The company also launched Muse Image, an Instagram image generator that uses public accounts' photos by default without notifying them. These moves reveal a pattern of prioritizing data collection over privacy, with the LED fix serving as a liability shield while the always-on mode and Muse Image expand Meta's data pipeline.", "body_md": "A\n\n[TechCrunch headline](https://techcrunch.com/2026/07/08/meta-wants-its-ai-glasses-to-seem-less-creepy-its-ai-strategy-says-otherwise/) crossed my Mastodon feed today about Meta wanting its AI glasses to \"seem less creepy,\" and I laughed reading it. Not because the premise is wrong. Because the timing is so bad it's almost generous of Meta to hand it to us like this.\n\nHere's the sequence. Meta announced its AI glasses will now disable the camera entirely if someone tampers with or covers the little white capture LED — the light that's supposed to tell bystanders \"hey, this thing might be recording you.\" Good feature, on its face. People have been\n\n[figuring out how to defeat that LED](https://www.digitaltrends.com/wearables/meta-could-be-working-on-always-seeing-always-hearing-ai-smart-glasses-with-ai-super-sensing-power/) since the glasses launched, and closing that hole is a reasonable thing to ship.\n\nExcept in the same week, the Financial Times reported Meta is testing a \"super-sensing\" prototype that would keep the camera and mic running continuously for hours, snapping photos every few seconds. And internally, Mark Zuckerberg has reportedly\n\n[questioned whether the LED needs to stay on at all](https://aiweekly.co/alerts/meta-tests-always-on-super-sensing-mode-for-next-ray-bans) in that always-on mode. So the company patched the one way you could sneakily defeat the recording light, the same week it's debating whether to just turn the light off by design instead.\n\nThat's not a security fix. That's negotiating with yourself over which door to leave open.\n\nAnd it raises the question Meta never seems to actually ask itself: if people have moved from taping over the LED to, in Meta's own words, \"sophisticated efforts to modify or destroy\" it, at what point does the answer stop being \"patch this specific workaround\" and start being \"maybe don't sell a covert recording device\"? That's not a rhetorical jab — it's the obvious next question once you see how many rounds of whack-a-mole this has already gone through. Every fix preserves the feature and the data pipeline behind it; none of them ask whether the feature should exist in its current form at all. A company that actually prioritized bystander consent over data collection has a much simpler option on the table: disable always-on capture rather than harden it against tampering.\n\nIt doesn't stop at hardware, either. The same week, Meta launched Muse Image, its new in-house image generator, baked directly into Instagram. The headline feature: you can @-mention any public Instagram account in a prompt, and Meta AI will pull that person's actual photos to generate a new image using their likeness. For public accounts, this is\n\n[on by default](https://www.macrumors.com/2026/07/08/meta-ai-muse-image/) — you have to go find the toggle yourself, buried under Sharing and reuse, and switch off Posts and Reels separately. Meta's own help page confirms you won't even be notified if someone does this to you.\n\nThree privacy stories, one week, and only one of them was allowed to look like a privacy win.\n\nI know, I know. Meta does something shady — shocking! The company's entire business model runs on harvesting as much of the planet's data as it can and turning it into ad targeting, and everyone reading this already knows that. I'm not trying to relitigate that. What's worth sitting with here isn't the general \"Meta bad\" take, it's the specific mechanics of this one week: which feature got the press release, which one got buried, and how neatly they line up with what actually costs the company money versus what actually makes it money.\n\nI don't think this is Meta losing the thread. I think it's the plan working exactly as intended. Look at what each piece actually protects. The anti-tamper LED feature costs Meta nothing in data, and it shields Meta from liability the moment an individual user secretly records someone — the wiretapping statutes, the more-than-a-dozen states with all-party consent laws. If some rogue user gets caught recording someone without consent, Meta wants to point at the tamper-proof LED and say that's on the user, not us. The always-on glasses are the actual roadmap. Ambient sensor data — what you're looking at, what's being said around you, who's in the room — is the one category of data Meta has never had at platform scale, and it's a fundamentally more valuable dataset than anything the social graph ever gave them. And the Instagram opt-out exists so Meta can tell whichever regulator asks that users had a choice. A\n\n[privacy tracker guide](https://www.trendingtopics.eu/how-to-stop-meta-ai-from-processing-your-instagram-content/) on this spells out the quiet part directly: if Meta only unlocked the feature after active consent, the pool of usable content would stay small. That's the whole game, stated plainly by someone writing a how-to.\n\nWhatever protects Meta from getting sued or fined ships loud, with a press release. Whatever expands what Meta can collect ships quiet, defaulted on, buried three menus deep.\n\nThe thing that got me digging further is how identical the shape is over at Google, even though the vibe there is completely different. Google doesn't have Meta's \"creep's weapon\" reputation to manage, so it doesn't need the theatrical privacy win. But underneath, it's running the same illusion-of-choice architecture.\n\nGoogle rolled Gemini into Gmail, Chrome, and Workspace by default for US users, and only opted out EU, UK, and Japan users because GDPR forced its hand. A\n\n[class-action suit](https://felloai.com/how-to-turn-off-google-gemini/) filed in November accuses Google of quietly flipping the US default around October 2025, with the opt-out buried three menu layers deep — and even then, disabling \"Smart features\" in Gmail also kills spellcheck, grammar suggestions, and Smart Compose, because Google never unbundled the AI toggle from features people actually want. Privacy researchers are calling that bundling a dark pattern outright, and it's hard to argue otherwise. Turn off \"Gemini Apps Activity\" and Google will still hold your conversation for 72 hours \"for operational purposes,\" according to a\n\n[privacy-focused write-up](https://pixelunion.eu/blog/google-photos-and-gemini/) covering the policy. Deletion isn't instant even when you explicitly ask. In January, Google launched Personal Intelligence, letting Gemini reason across your Gmail, Photos, YouTube history, and Search in one shared context. It's opt-in for now, but as one\n\n[enterprise security analysis](https://concentric.ai/google-gemini-security-risks/) put it, what's a premium opt-in beta today tends to become default infrastructure within eighteen months. Nobody's accusing Google of lying about its current settings. The concern is what the default looks like once the feature graduates out of beta.\n\nA recent Epic-backed\n\n[dark patterns study covered by Gizmodo](https://gizmodo.com/the-dark-patterns-keeping-you-from-opting-out-of-data-sharing-are-darker-than-ever-study-finds-2000761379) found more than a dozen major platforms — Meta, Google, and OpenAI among them — don't even clearly link to their own opt-out forms from their homepage or privacy policy, despite 21 states now legally requiring an easy one. Same underlying move as Meta, just with better manners. Ship the feature broadly capturing data, put the actual control behind friction, and save the loud \"we respect your privacy\" messaging for whatever costs the least. Google's just slicker about it. No camera-glasses controversy to manage means the theater can stay quiet and let the defaults do the talking instead.\n\nWhich loops back to the question nobody at Meta seems willing to sit with: they've now confirmed, in their own words, that this is an arms race against their own users — and the response was another round of patching, timed to the same week they were weighing whether to remove the warning light altogether. That's the tell. A company actually worried about bystanders picks disable. A company worried about the dataset picks patch.\n\nLet's be clear: there are some legitimate use cases for both AI and smart glasses. Nowhere near as many as Meta, OpenAI, and the others would have us believe, but they exist. What it really means is to know what you're using and what it costs. Cost isn't just the $20/month or whatever. Cost is also what you hand them in data. Read the mechanics, not the press release. The feature getting the headline this week is rarely the feature that matters next quarter.", "url": "https://wpnews.pro/news/always-on-always-watching", "canonical_source": "https://blog.ppb1701.com/always-on-always-watching", "published_at": "2026-07-09 13:38:26+00:00", "updated_at": "2026-07-09 13:49:42.045538+00:00", "lang": "en", "topics": ["ai-ethics", "ai-products", "ai-policy"], "entities": ["Meta", "Mark Zuckerberg", "Ray-Ban", "Instagram", "Muse Image", "Financial Times", "TechCrunch"], "alternates": {"html": "https://wpnews.pro/news/always-on-always-watching", "markdown": "https://wpnews.pro/news/always-on-always-watching.md", "text": "https://wpnews.pro/news/always-on-always-watching.txt", "jsonld": "https://wpnews.pro/news/always-on-always-watching.jsonld"}}