September 16, 2026
TechCrunch reported that Meta is preparing to sell a version of its smart glasses without a camera, following accusations that its camera-equipped models amounted to 'perv glasses' -- devices that let wearers secretly record people without their knowledge. On paper, removing the lens solves the immediate problem: no camera, no covert recording. But I don't think this is the fix Meta's PR team hopes it is. The backlash was never really about one sensor. It was about the fact that Meta shipped a wearable capable of discreet surveillance and only now, after the accusations, is offering an option that dials it back. That's a company reacting to a controversy, not one that thought through consent and bystander privacy before launch.
For business buyers eyeing smart glasses for field service, retail, or hands-free workflows, this matters beyond the tabloid framing. A camera-free SKU might satisfy privacy-conscious customers or workplaces with strict recording policies, but it also strips out the exact feature that made the glasses useful for many enterprise use cases -- visual documentation, remote assistance, quality checks. Meta is essentially offering two products for two audiences: one that's more trustworthy and one that's more useful, and asking buyers to pick a side. That's a tell that the underlying design didn't have a good answer for both. Any vendor selling AI-connected hardware to your team should be able to explain, clearly, who can see what data and when -- the kind of governance question we've written about in the context of AI agent permissions, which applies just as much to a camera on someone's face as it does to software with access to your CRM.</p>
Separately, TechCrunch reported that SK Hynix is in talks with Intel about building memory chips in the US, though SK Hynix told TechCrunch that nothing is finalized. If this goes through, it's a notable shift: a South Korean memory giant partnering with an American chipmaker on US soil, at a moment when every AI company on earth is fighting over memory and compute capacity. Memory chips aren't the flashy part of the AI story, but they're the bottleneck that determines how fast and how cheaply everyone else -- from OpenAI to the startup building your customer-service chatbot -- can actually run their models.
I read this less as a headline about one deal and more as a signal about where the entire industry is heading: toward domestic, redundant supply chains for the physical guts of AI, because nobody wants to be caught short again. For a small or mid-sized business, none of this shows up on an invoice tomorrow. But it's exactly the kind of upstream shift that eventually shows up as pricing and availability for the AI tools you rely on. It's a good reminder that the tools sitting on top of this infrastructure -- the software you actually buy -- are worth evaluating on their own terms, independent of which chip vendor wins this particular negotiation. That's part of why we keep pushing the buy vs. build vs. ViibeStack comparison: the underlying hardware race is out of your control, but your software stack doesn't have to be hostage to it.
Meta's glasses and SK Hynix's chip talks look unrelated, but both are stories about decisions made far upstream of the end user that quietly shape what you're allowed to trust and what you're able to afford. Meta decided what a camera on your face should be capable of before regulators or customers weighed in. SK Hynix and Intel are deciding, in a boardroom you'll never see, how much memory the next generation of AI tools will have to work with. Business leaders don't get a vote in either room, but they do get to choose which vendors they hand data and budget to -- and that choice matters more when the ground underneath is shifting this much.
If your business were offered a version of a tool that's more private but less capable, versus one that's more capable but riskier, which would you actually choose -- and would that decision survive contact with your actual customers?
Sources