Welcome to the latest edition of the AI Update, where we continue to track the major developments in artificial intelligence and their implications for identity security and cybersecurity more broadly.
Anthropic and the White House are fighting again. Just days after launching its first ‘Mythos-class’ AI model, Fable 5, Anthropic was hit with an export-control order demanding that Anthropic prevent the model from being accessed “by any foreign national, whether inside or outside the United States.” Anthropic, in turn, told its customers that the only practical way to comply would be to pull the model from public access, which it went ahead and did. Its team has been negotiating with the White House in the week since. At the time of writing, they still haven’t reached an agreement, but there have been positive signs that a deal may be in the works.
There are a couple of funny things about that situation. One is that Anthropic has been perhaps the loudest voice among the frontier AI labs calling for government regulation, and now it’s being regulated in the most heavy-handed way you could expect. The other is that Amazon CEO Andy Jassy is reportedly the person who set the whole thing in motion, by warning the White House about potential security risks in Anthropic’s new models. Amazon is, of course, one of Anthropic’s AI rivals.
Some OpenAI financials have reportedly leaked, showing that in Q1 of this year the company spent $3.7 billion, while revenues came in at $5.7 billion. Last year, OpenAI reportedly saw $13 billion in revenue with $34 billion in total costs and expenses. The leaked data also show that OpenAI ended the latest quarter with over $73 billion in cash, up from $40 billion at the end of 2025; and that it’s planning to spend about $32 billion this year on training its future AI models. Next year, OpenAI expects that number to rise to $65 billion. This all comes as OpenAI aims for an IPO later this year.
Midjourney has announced a new full-body ultrasonic body scanner, in one of the most surprising turns in the AI space so far. Midjourney has become renowned for an image and video generation platform that offers particularly striking aesthetics, winning a loyal customer base. Now, somehow, it’s launching a subsidiary called “Midjourney Medical” to focus on its body scanner tech. The scanner has a platform that lowers a patient into a shallow body of water, through which an array of underwater sensors create an MRI-like body scan, purportedly in just one minute. It’s going to launch at a planned health “spa” in San Francisco next year, and Midjourney’s management are aiming to eventually open thousands of spas across the country.
A new anti-deepfake bill has moved to the U.S. Senate. The NO FAKES Act – short for Nurture Originals, Foster Art, and Keep Entertainment Safe – would give people intellectual property-style rights to their own voice and likeness, allowing them to take action against unauthorized AI-generated “digital replicas” in audiovisual works or sound recordings. It would also require platforms to remove unauthorized deepfakes when they know the content was not licensed, and would create liability for distributing unauthorized deepfakes or offering products primarily designed to make them. Supporters include entertainment and creator groups such as SAG-AFTRA, the Recording Academy, Universal Music Group, Warner Music Group, Sony Music, Disney, CAA, WME, the Motion Picture Association, the Television Academy, ASCAP, BMI, and the Authors Guild, as well as tech or platform companies including YouTube, TikTok, IBM, OpenAI, Hive, and Vermillio.
Graham Norton, the host of a popular British talk show, has secured a California court order compelling Meta to provide information that could help identify the owner of a Facebook page that had published misleading and defamatory deepfakes about him and his husband. Norton told the court he plans to use the information from Meta to pursue legal action in England, because English courts could not directly reach the identifying records held by Meta in the U.S.
The chatbot’s take: We wanted more info about Graham Norton, whose boozy talk show has produced some truly delightful interview clips.
–
By Alex Perala
Follow Us