Pentagon deploys ChatGPT and Grok, Instagram cracks down on AI influencers, and Nvidia's $3.5B chip bet

Picture the US military handing every soldier a login to ChatGPT. That moment arrived this week - and it happened quietly, tucked inside a broader story about how AI is reshaping power, platforms, and the chips that run everything. Today's newsletter connects some dots that I think matter more than they look at first glance.

Under the EU's Digital Services Act, a platform can be classified as a 'Very Large Online Platform' or a 'Very Large Online Search Engine'. What monthly active user threshold triggers this designation?

The answer is hiding near the bottom of today's newsletter... keep scrolling. 👇

The US Department of Defense has officially added versions of OpenAI's ChatGPT and xAI's Grok to its central AI portal - joining Google's Gemini, which was already available. Pentagon personnel now have a single interface where they can access multiple frontier models side by side, something no other government has formally announced at this scale.

These aren't vanilla consumer versions - they're adapted builds designed to work within defense security requirements. The move signals that the US military is treating large language model access as standard infrastructure, not a niche experiment. It also raises immediate questions about how classified or sensitive queries will be handled, and whether different models will be approved for different clearance levels.

The Pentagon treating ChatGPT and Grok like standard-issue tools - alongside Gemini - marks a genuine shift in military AI adoption, and it's a sign that frontier model access is becoming baseline infrastructure across major institutions, not just tech companies.

Read the full story →

Instagram is putting new limits on AI accounts that have been passing themselves off as human creators. Profiles featuring AI-generated people - think Aitana Lopez and similar synthetic influencers - will now display an "AI-generated profile" label, replacing the previous "AI creator" tag that most users scrolled straight past. On top of that, undisclosed AI profiles will see their reach throttled.

The timing isn't coincidental. Frustration over AI influencers has been building for months, and Instagram's parent company Meta is under pressure from regulators and users alike. The label change is subtle but meaningful - 'AI creator' could mean a human using AI tools, while 'AI-generated profile' makes it explicit that the entire person is synthetic.

The reach restrictions on undisclosed AI accounts are arguably the bigger lever. Labels can be ignored. Algorithmic suppression hits creators where it hurts - in engagement and revenue. This is Meta acknowledging that voluntary disclosure isn't working, and deciding to enforce it through the feed itself.

If you manage brand partnerships or influencer campaigns, this changes your vetting process - verifying whether an account is AI-generated is now officially Instagram's problem too, but the responsibility to check won't disappear entirely from marketers.

Read the full story →

Here's the threat Nvidia is navigating: Google, Amazon, Microsoft, and Meta are all developing custom AI chips in-house. If that trend accelerates, the world's biggest AI spenders stop buying Nvidia GPUs and start making their own. Nvidia's $3.5 billion investment in Taiwanese chipmaker MediaTek is its answer to that threat.

The strategy: by partnering deeply with MediaTek, Nvidia positions itself as the design and IP layer even when the underlying silicon isn't its own. Rather than competing head-on with custom chips, it embeds itself into the supply chain that those chips depend on. MediaTek's scale - especially across edge and mobile devices - gives Nvidia reach into markets where its flagship GPUs are too power-hungry and expensive to land.

This also connects to a broader Nvidia pivot toward AI infrastructure and networking - not just raw GPU power. By staying essential to the plumbing of AI compute rather than just the processing, Nvidia makes itself harder to cut out even as Big Tech diversifies its silicon stack.

This deal is Nvidia hedging against a world where its best customers stop being its best customers for raw compute - and making sure it still gets paid either way.

Read the full story →

The European Commission has officially designated ChatGPT as a Very Large Online Search Engine under the EU's Digital Services Act - meaning OpenAI now faces the same category of regulation as Google Search and Bing. The DSA threshold for this designation is 45 million monthly active users across the EU, and ChatGPT has clearly cleared that bar.

What does this actually mean in practice? OpenAI will be required to demonstrate how it mitigates risks related to ChatGPT's impact on minors, user mental health, and the spread of illegal content. It also opens the door to audits, transparency reporting requirements, and significant fines for non-compliance - potentially up to 6% of global annual turnover.

The classification is a meaningful escalation. For years, ChatGPT existed in a regulatory grey zone - it wasn't quite a search engine, wasn't quite a social platform. Now the EU has made a call, and AI regulation frameworks are starting to catch up with how people actually use these tools. If you're building products on top of ChatGPT that serve EU users, the compliance picture just got more complex.

Being classified as a search engine under the DSA is a big deal - it hands regulators a clear legal hook to scrutinize ChatGPT's outputs, its recommendations, and its handling of vulnerable users in ways that weren't available before.

Read the full story →

Apple says it has "shocking evidence" that a former employee destroyed data after learning he was under investigation for allegedly stealing company secrets and passing them to OpenAI. The case is escalating fast: Apple claims the individual deleted files and wiped devices once he realized investigators were closing in - a move that, if proven, turns a trade secrets case into something considerably more serious.

The specific nature of what was allegedly stolen hasn't been fully disclosed publicly, but Apple has described the data as significant and company-sensitive. This isn't just an internal HR story - it sits at the intersection of corporate espionage, AI development pipelines, and the increasingly fierce competition for proprietary research and engineering talent between the biggest players in tech.

The case is a reminder that the race to build better AI isn't only playing out in benchmarks and model releases - it's also happening in courtrooms, and the legal battles are getting messier as the stakes climb higher.

Read the full story →

The answer is 45 million users! Under the EU's Digital Services Act, any platform or search engine with more than 45 million monthly active users in the EU is designated as 'very large' - triggering the heaviest tier of compliance obligations, including mandatory risk assessments, independent audits, and potential fines of up to 6% of global turnover. ChatGPT has now officially crossed that line.

By the way - if you're building a web presence for your AI projects or side businesses, check out 60sec.site - an AI website builder that gets you live in under a minute. No coding required, no templates to wrestle with. Worth a look if you've been putting off launching something.

The Pentagon now has ChatGPT, Grok, and Gemini available to its personnel through one portal. Do you think giving military staff access to frontier AI models is a smart move - or does it worry you? Hit reply and give me your honest take. I read every response, and this one I'm genuinely curious about.

That's it for today - see you tomorrow with more from the fast-moving world of AI. And if you want to browse past coverage, the full Daily Inference archive is always there when you need it.