Three stories crossed my desk today and they all point at the same thing: the AI money and the AI trouble are showing up in the same places now. Let's get into it.
Nvidia Wants a Piece of Perplexity
Word is Nvidia is in talks to put equity money into Perplexity at a valuation north of $30 billion. Perplexity's annualized revenue has reportedly crossed $750 million, which is real growth for a company that started as "what if search but the robot writes the answer."
Why it matters: Nvidia keeps doing this. They chip in cash to the companies that are about to go buy a whole lot of Nvidia chips. It is not charity, it is a flywheel, and it keeps Nvidia's fingerprints on basically every layer of the AI stack at once, from silicon to the app you actually open on your phone.
Robert's take: I like Perplexity fine, I use it some weeks more than Google. But a $30 billion tag on a company still figuring out how to make search-as-answers actually profitable tells you this market is pricing in a whole lot of hope. Nvidia is not investing because they think Perplexity has cracked the business model. They are investing because they need somebody besides OpenAI and Google buying GPUs at that scale. Watch the chip orders, not the press release.
OpenAI Shuts Down a Russian Misinformation Op
OpenAI pulled the plug on a batch of ChatGPT accounts it says were tied to a covert Russian influence campaign, using the model to help pump out fake content for an online misinformation push.
Why it matters: this is not the first time and it will not be the last. Every big model provider is now, whether they signed up for it or not, in the business of counter-intelligence. Detecting coordinated abuse at scale is a real technical problem, and it is only getting harder as the tools get cheaper and more capable.
Robert's take: good on OpenAI for catching it and saying so publicly instead of quietly banning and staying quiet. But let's be honest, for every operation they catch there are probably a few they have not found yet, running on OpenAI's stuff or somebody else's. If you build a machine that writes convincing text on command, you signed up to also build the machine that watches for people abusing it. That second machine needs way more investment than it's getting right now, industry wide.
Alabama's AG Comes After OpenAI on Safety
Alabama Attorney General Steve Marshall issued a subpoena to OpenAI, digging into safety risks tied to autonomous AI models. This is a state law enforcement office asking hard questions about what happens when these systems act on their own without a human checking every step.
Why it matters: this is state AGs testing whether they can regulate AI safety the same way they've gone after other industries, one subpoena and one state at a time, instead of waiting on Congress to pass something. If Alabama gets traction, expect a line of other states behind them.
Robert's take: I have said it before, the patchwork of state-by-state AI rules is coming whether the labs like it or not, because nobody trusts Washington to move fast enough. That is messy for companies trying to build one product for the whole country, but it is also what happens when an industry moves faster than the guardrails. OpenAI and everybody else in this race should have been more forthcoming on their own instead of waiting to get subpoenaed into it.
The Thread Connecting All Three
Big money keeps flowing into AI like Perplexity's raise, real abuse keeps showing up like the Russian accounts, and real scrutiny keeps landing like Alabama's subpoena. All three happened inside about 24 hours. That is where we are now. The industry is not choosing between growth and accountability anymore, both are just going to keep happening at the same time, and every company in this space better get comfortable moving in that kind of weather.