AI2Day Daily Brief — 14 Aug 2026

Good morning, it is Friday the fourteenth of August, and the big story today is a full-length feature film made by a crew of eleven people using AI tools, and it is already online for anyone to watch.

Higgsfield, an AI filmmaking platform, has released Cully Hill Boys, an action comedy about three friends in a London pub who stumble into a crime lord's money. Three public figures star in it, streamer N3on, content creator Matt Kiatipis, and UFC fighter Israel Adesanya, all under signed licensing deals, none of whom shot a single scene. The script came from a human writer, Tim Planagan, who kept his rights. Anthropic's Claude generated text prompts, while tools from ByteDance and Google handled visuals. Higgsfield also published most of the prompts used, making this as much a public demonstration of the technology as it is a film.

Transparent of them. Next, money, a lot of it.

Anthropic investors are telling the Financial Times they expect the company to go public at a valuation of two trillion dollars or more, most likely this October. Six backers shared that view, and if the number holds it would make the IPO, the moment a private company first sells shares to the public, the largest in history. It would also more than double Anthropic's current private valuation and put it ahead of SpaceX. Anthropic is five years old and best known for building the Claude family of AI assistants, which compete directly with OpenAI's ChatGPT. The company says rapidly rising revenue supports the figure, though public markets are already showing nerves about AI valuations, so the debut will be a real test.

A two-trillion-dollar question. Speaking of numbers being tested, story three is about AI research and how much of it actually holds up.

Between the fifteenth of July and the second of August this year, more than twelve hundred volunteers attempted to reproduce the scientific claims in over two thousand two hundred papers presented at the International Conference on Machine Learning. They used coding agents, software that can carry out multi-step tasks automatically, to run the experiments and check the results. Of the papers reviewed, fifty-one percent had at least one claim independently verified. But twenty-three percent had at least one claim falsified or contested. In two hundred and forty-two papers, different teams reached opposite conclusions on the same claims. The project was first reported by Hugging Face, and it is a useful reminder that a published result and a confirmed result are not the same thing.

Worth keeping in mind. Now to OpenAI, which is heading toward its own public listing and just lost two senior executives in the same week.

Denise Dresser, OpenAI's Chief Revenue Officer since December last year, announced on the thirteenth of August that she will leave in the coming weeks. That came just two days after Brad Lightcap, the former Chief Operating Officer, said he was also leaving. At least four senior executives have departed OpenAI in recent months. The company filed a confidential registration document with the Securities and Exchange Commission in June, the first formal step toward a public stock offering. Dali Rajic, currently president and chief operating officer at cloud-security firm Wiz, will step into the revenue chief role. OpenAI has not commented on the pattern of departures, and the timing, with an IPO on the horizon, makes each exit harder to ignore.

Are you at risk? It takes one wrong click, on one bad email, to bring a whole company down. Train2Secure turns your staff from your biggest risk into your strongest defence, with realistic phishing simulations and quick security training that actually sticks. From $1.59 per user, per month, cheaper than a single cup of coffee. Start your free trial at Train2Secure dot com. That's Train, the number two, Secure, dot com.

A lot of movement at the top. Last story, and it is about who actually sells AI to big companies at scale.

IBM and OpenAI announced a partnership on Thursday to jointly market AI products to large enterprise customers in financial services, government, telecommunications, and retail. IBM plans to train and certify tens of thousands of its consultants on OpenAI technologies within the next several months, mostly by retraining people already on the payroll. OpenAI's latest models, including GPT five point six and Codex, its coding-focused tool, will be added to IBM's internal consulting platform. This follows a similar alliance IBM struck with Anthropic less than a year ago. IBM says it is not locking into one provider, and the strategy is straightforward: let others build the models, then use IBM's existing relationships with the world's largest organisations to put those models to work. Financial terms were not disclosed. First reported by TechCrunch AI.

For a deeper look at all five of these stories and more, join us on Monday for AI Today Weekly. That is your AI briefing for today. Every story is at A-I-2-Day dot live. That is A, I, the number two, D-A-Y, dot live. We are back tomorrow morning. If you got something out of this, a thumbs up and a subscribe genuinely helps.