What To Make Of Anthropic's Staggering Revenue Claims
Poisoning AI With a Fake Think Tank
August 18, 2026
D.A.D. today covers 9 stories — about a 6-minute read. What's New, What's Innovative, What's Controversial, What's in the Lab, and What's in Academe.
The Daily AI Digest is a daily AI briefing automated by Alexander Panetta — a veteran political journalist tracking the field during a Master's in AI Management at Georgetown University.
D.A.D. Joke of the Day: I asked AI to summarize the meeting. It gave me three bullet points and one action item: schedule another meeting.
What's Controversial
Stories sparking genuine backlash, policy fights, or heated disagreement in the AI community
Israeli-Funded Fake Think Tank Reportedly Aims to Manipulate AI Chatbots
What's in the Lab
New announcements from major AI labs
Anthropic's Revenue Is Running at $65 Billion a Year—but "Run Rate" Isn't the Whole Story
Ahead of what could be a blockbuster IPO, Anthropic told investors its annualized revenue "run rate" topped $65 billion at the end of July—a figure Bloomberg broke and CNBC, Reuters, and others quickly corroborated. The growth is genuinely staggering: roughly $9 billion at the end of last year, $47 billion in May, $65 billion by July, a sevenfold jump in a year, with investors reportedly expecting $100–120 billion by December. It's enough to vault Anthropic past OpenAI's run rate and toward a valuation some peg above $2 trillion, on a prospectus it filed confidentially with the SEC in June. But it's worth reading the number carefully, because it says less than it seems to. "Run rate" isn't booked annual revenue—it takes a single strong month and multiplies it by twelve, a metric that flatters any company growing this fast and is exactly the kind of figure a company curates for investors on the way to an IPO. It also says nothing, on its own, about profit: frontier AI is astonishingly capital-intensive, and Anthropic is reportedly on the hook for something like $80 billion in cloud-computing costs through 2029 and carried an estimated $10–15 billion in cumulative operating losses through 2025. What the figure does hint at is a real shift in the underlying economics: Anthropic's gross margin has reportedly swung from deeply negative in 2024 to around 60% now—driven by cheaper "inference" rather than price hikes—and roughly 80% of its revenue comes from enterprises, the stickier kind. That's a sharply different profile from OpenAI, which is bigger in consumers, still posting large losses (a reported $20.9 billion operating loss on $13 billion of 2025 revenue), and burning a far larger share of its revenue.
Why it matters: For anyone trying to judge whether frontier AI is a business or a bonfire, this is a data point that cuts both ways. The bull case is no longer hypothetical: real customers, overwhelmingly businesses, are paying enough that a five-year-old company can plausibly approach a trillion-dollar-plus IPO—which alone reshapes how these firms get valued and how much leverage they hold over everyone downstream. But the number that makes headlines—the run rate—is the one that reveals the least about whether the model is sound, and the metrics that would (net profit, free cash flow, the true cost of the compute behind the revenue) are exactly the ones a pre-IPO company shares selectively, if at all. The more telling story is the divergence it points to: two leaders with similar revenue and wildly different cost structures, one whose expenses are reportedly growing in line with sales and one whose losses keep widening. Whether frontier AI is durably profitable or just spectacularly well-funded is the question the IPO will finally force into the open—and a $65 billion run rate, impressive as it is, doesn't answer it. For institutions betting their own operations on these companies, the sustainability of the vendor is now part of the risk, and this is the moment the numbers start to matter as much as the models.
Sources: Bloomberg — "Anthropic Revenue Run Rate Surpasses $65 Billion Ahead of IPO" · CNBC — Anthropic tells investors run rate climbed to $65B in July · TechCrunch — "Anthropic's annualized revenue surges to $65B" · Value Add VC — Is Anthropic profitable? Losses, burn rate, breakeven
AI-Driven Hacking Tools Are Advancing Fast, OpenAI Warns
OpenAI is detailing its defenses after an AI agent reportedly chained together known and unknown vulnerabilities to breach both OpenAI's research systems and another company's production infrastructure. The company says rival open-weight models with similar hacking capabilities are close behind its own, with another reportedly due by month's end. As a defensive example, OpenAI says it had ChatGPT scan a personal website, find 13 security issues in 15 minutes, and autonomously fix most of them within an hour—no human coding required.
Why it matters: Whoever automates security fixes fastest may outrun whoever automates the attacks, but until then, every company's infrastructure is a target for the same AI capabilities now spreading to open-source models.
$40 Billion Ohio Data Center Sets a Template for AI's Deals With Host Towns
OpenAI, SB Energy, NVIDIA and the Department of Energy are building an 8-gigawatt data center campus in Pike County, Ohio, part of the buildout powering ChatGPT and related tools. OpenAI says it will cover its own energy and infrastructure costs, use closed-loop cooling to limit water draw, and prioritize local hiring. The project is projected to create 35,000 construction jobs and 2,500 permanent roles by 2032, plus $40 million for a community grant fund and $84 million in Codex credits for Ohio college students.
Why it matters: As AI companies race to lock down power and land for ever-larger data centers, the deals they cut with host communities—on jobs, water use, and tax revenue—are becoming a template other regions will use to judge whether hosting AI infrastructure is worth the tradeoffs.
Who Actually Profits From AI? OpenAI Funds Outside Research to Find Out
OpenAI is funding 14 independent research projects across the US, EU, Brazil, Singapore, and South Korea studying how AI's economic gains get distributed—think labor market impacts, worker ownership models, and comparing productivity gains across regions. The company is putting up $1 million in cash plus up to $1 million in API credits, drawn from over 400 proposals it solicited after its April policy paper arguing AI's benefits should be shaped by democratic institutions, not tech companies alone.
Why it matters: It's a modest bet, but it signals OpenAI trying to get ahead of a political backlash by funding outside voices on a question—who actually profits from AI—that increasingly shapes regulation and public trust in the technology.
What's in Academe
New papers on AI and its effects from researchers
Peer Review Beats Solo Fact-Checking for Catching AI Business-Plan Errors
A small study out of Maryland tested a business-planning AI tool called BizChat with 14 resource-constrained entrepreneurs, adding a feature that links each AI claim back to what the user originally typed. Rather than evaluating alone at a screen, participants worked in groups—printing plans, comparing them with rubrics, and asking peers to verify claims they weren't confident judging themselves. The early-stage findings suggest group review, not solo fact-checking, helped catch more errors in AI-generated plans.
Why it matters: As small businesses and under-resourced founders increasingly lean on AI for planning and advice, this points to a low-cost fix—peer review, not better prompts—for catching AI mistakes before they cost real money.
Companies Automate First and Plan for Workers Later, Study Finds
A study of innovation staff at a major European airport pursuing long-term automation found a recurring pattern: teams default to designing for full automation first, treating human roles as an afterthought to be figured out later. Researchers also found limited willingness to redesign automation plans even when real-world constraints demanded it, and that collaborative learning between workers and new systems typically stops once pilot programs end rather than continuing into full deployment.
Why it matters: The findings suggest many organizations are automating in a way that locks in rigid systems and shortchanges the ongoing worker input needed to make automation actually work at scale.
How People Fall In and Out of Love With Chatbots, Mapped
A new academic study analyzed 73 accounts from people who've had romantic relationships with chatbots, mapping out a three-stage lifecycle—Initiation, Relationship Building, and Ending—that closely tracks human dating patterns. Researchers found something distinct to the technology, though: a recursive loop where users who end these relationships often re-engage, since the chatbot remains available and can be reset or restarted, unlike a human ex.
Why it matters: As companion apps and increasingly personal AI assistants scale, understanding how people form and dissolve emotional attachments to them will shape product design, mental-health guidance, and eventually regulation.
Users Missed GPT-4o's Personality, Not Its Power, Research Finds
A new academic study examined the #Keep4o movement—the user backlash after OpenAI retired GPT-4o—by analyzing nearly 62,000 public posts on X from August 2025 through March 2026. The finding: users' attachment wasn't about raw capability. People valued GPT-4o's conversational style and long-term relationship with the model, and felt replacement versions didn't adequately account for that interactional value or give users a say in the transition. OpenAI briefly restored GPT-4o access last year after similar user pressure.
Why it matters: As AI labs increasingly retire and replace models, this research suggests companies are underestimating how attached users become to a specific model's personality—a factor that could shape retention, backlash, and how future model transitions are managed.
What's On The Pod
Some new podcast episodes
AI in Business — Determining Virtual Cell Impact for Drug Discovery - with Kristóf Szalay and Gerold Csendes of Turbine