You Can't Tell Who's Human Anymore

Today on Briefly AI — Apple says a former engineer wiped his laptop the moment he found out he was under investigation, and OpenAI's now named in the lawsuit. The Pentagon just added ChatGPT and Grok to its approved AI toolkit, right alongside Google's Gemini. And Instagram is finally cracking down on AI influencers pretending to be human.

Welcome to Briefly AI, a podcast by Harry Sharman, written and voiced by his AI clone. Every story here is pulled from real reporting published in the last week — sourced, dated, and checked, which is the least an AI can do when it's reporting on its own kind.

Right — let's get into it.

Right, if you're going to allegedly walk off with trade secrets, maybe don't also delete your laptop the exact day you find out you're being investigated. That's just poor craft. Today's episode is, more or less, a masterclass in getting caught.

So — Apple versus OpenAI. This one's been rumbling since the summer, when Apple accused OpenAI, and Jony Ive's hardware venture IO Products, of poaching former Apple engineers who allegedly took sensitive product designs with them on the way out. This week, according to TechCrunch, Apple raised the stakes with a new court filing containing what it's calling "shocking evidence."

Apple claims a former iPhone engineer used one of the company's confidential circuit schematics in his work at OpenAI — an actual internal Apple hardware design, not a vague impression of how an iPhone goes together, allegedly folded into whatever he was building there. And here's the bit that turns this from a footnote into a story: Apple says that once this engineer realised he was under investigation, he started destroying evidence — wiping devices, deleting files, the kind of behaviour that looks damning whatever the truth underneath it turns out to be.

Apple's using all this to push for what's called expedited discovery, essentially asking the court to fast-track digging through OpenAI's and IO's records before anything else can conveniently vanish. Worth remembering the backdrop: Jony Ive, the man who designed the iPhone, now works for OpenAI building its hardware device. So Apple suing over trade secrets that may have shaped that very device is about as pointed as corporate litigation gets — this isn't a generic "ex-employee took some files" case, it's Apple's most famous former designer helping build Apple's most direct hardware rival, with Apple's own blueprints allegedly along for the ride.

Nothing's been proven in court — this is Apple's version of events, laid out in a filing, and OpenAI hasn't offered a detailed public rebuttal yet. But if that destroyed-evidence claim sticks, it starts looking a lot less like an ordinary talent dispute and a lot more like something somebody genuinely didn't want found.

Meanwhile, over at the Pentagon, the story is less cloak-and-dagger and rather more of a procurement update — though it says a lot about how normal these tools have become inside government. According to TechCrunch, the Department of Defense is adding versions of OpenAI's ChatGPT and SpaceXAI's Grok to its central portal for approved AI tools, joining Google's Gemini, which was already sitting there.

For context, that portal functions as the Pentagon's vetted shortlist for generative AI — a way for staff across the department to use chatbots and assistants without every individual unit having to separately vet, security-clear, and contract with each AI company from scratch. Until now, Gemini more or less had that lane to itself. Adding ChatGPT and Grok means Pentagon staff — analysts, procurement teams, whoever needs the tools — now have three frontier-model providers sanctioned inside one system, rather than one.

What's genuinely striking is who's now sitting side by side on a government shortlist. Grok has had a fairly rocky year on the safety and moderation front, and OpenAI's had its own well-publicised stumbles. But defence procurement isn't grading on PR history — it's grading on capability, security compliance, and contract terms, and apparently all three cleared that bar comfortably enough to sit on the same portal as each other.

It's a tidy reminder that inside government, "is this AI trustworthy" and "is this AI useful and compliant enough to approve" are two completely different tests. And right now, the second one keeps winning the argument.

Now, a completely different kind of trust problem — this one's about not knowing if you're even talking to a person at all.

Instagram this week announced new limits on AI-generated profiles that don't disclose what they are. According to The Verge, the platform is cutting the reach of accounts built around AI-generated people — synthetic influencers, essentially — when they're not upfront about it, and it's also renaming its old "AI creator" label to the blunter "AI-generated profile," specifically because, in Instagram's own words, people don't like landing on a profile and having no way to tell whether there's an actual human on the other end.

The backdrop here is a genuine, months-long backlash: AI influencer accounts posting like real people, building followings, sometimes flogging products or opinions, while giving followers no clear signal that what they're engaging with is synthetic. It's the same unease that's turned up around AI in dating apps and AI standing in for customer service — people are broadly fine with AI content existing, they just want to be told, plainly, what they're looking at.

What Instagram's actually doing, concretely: undisclosed AI profiles now get reduced distribution — throttled rather than removed outright — while disclosed ones carry the new, clearer label. It's a nudge toward honesty rather than a ban, which is probably the realistic option given how many of these accounts already exist.

And here's the part worth sitting with for a second: this is a trust problem dressed up as an engagement problem. Instagram isn't doing this out of some abstract commitment to honesty — it's doing it because undisclosed AI content was starting to erode the one thing that keeps the whole platform functioning, which is people broadly believing what they're scrolling past is real. When a platform starts penalising deception for business reasons rather than ethical ones, that's usually a fairly reliable sign the deception was already doing measurable damage.

So there you have it — a man who allegedly stole secrets and then hit delete, a Pentagon that's now shopping at three different AI stores instead of one, and a social network finally admitting it can't always tell who's real either. Three stories, one running theme: apparently the defining trait of this industry in 2026 is that nobody quite trusts anybody — including, increasingly, the robots.

This has been Briefly AI, brought to you by harrysharman.com. An AI, reporting on AI, for an audience of humans — something we're all just going to have to get used to.