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 teamed up with SpaceX this week in a huge compute agreement that will see Elon Musk’s AI company lease its entire Colossus 1 data center in Memphis, Tennessee, to one of its major rivals. (SpaceX acquired Elon’s xAI earlier this year and the team is looking to get data centers into orbit.) The deal notably comes in the second week of Elon’s lawsuit against OpenAI – Anthropic’s arch-rival – accusing its management of ‘stealing a charity’ by turning a startup Elon previously supported into a for-profit venture. “I spent a lot of time last week with senior members of the Anthropic team to understand what they do to ensure Claude is good for humanity and was impressed,” Elon said in a post on his X platform. “Everyone I met was highly competent and cared a great deal about doing the right thing. No one set off my evil detector.”
DeepSeek is reportedly close to raising about $45 billion in its first-ever external funding round. It’s notable in part for the state’s backing of an open source AI firm that has shown a strong ability to pursue the AI progress of closed-model Western rivals. China’s state-backed semiconductor investment vehicle, the China Integrated Circuit Industry Investment Fund – also known as the “Big Fund” – is reportedly in talks to lead the financing round. DeepSeek recently launched an advanced version of its flagship AI model and claimed state-of-the-art performance, though third-party assessments suggest it still trails leading closed models.
Arm says demand for its new Arm AGI CPU has already reached more than $2 billion across fiscal 2027 and 2028, more than double what the company had projected just weeks earlier when it first unveiled the chip. While Arm has made its name licensing chip designs, AGI CPU will be the first chip it actually builds and sells, and it’s aimed at “agentic AI” workloads. The catch is that Arm says it currently has enough supply to meet only about $1 billion of that expected demand, meaning the company’s early AI-chip opportunity is being constrained by manufacturing capacity, wafers, memory, and test equipment.
Peter Thiel is leading a $140 million Series B investment in Panthalassa, a Portland, Oregon-based renewable energy and ocean technology startup that wants to power AI inference at sea using wave energy. The idea is to have autonomous floating nodes that can generate electricity from wave motion, run AI chips onboard, use the ocean for cooling, and transmit inference results back to land by satellite, avoiding the need to bring offshore electricity back to the grid, and there are plans for “Ocean-3” deployments in the northern Pacific later this year and commercial systems in 2027. The round would reportedly value Panthalassa at nearly $1 billion.
EU governments and European Parliament negotiators have reached a provisional deal to simplify parts of the EU AI Act. Among other things, the provisional agreement would delay the enforcement deadline for “high-risk” AI systems from August 2 of this year to December 2, 2027. These systems include certain AI technologies in biometrics, critical infrastructure, law enforcement, employment, education, and access to essential services. The agreement also includes an EU-wide ban on AI-generated non-consensual sexually explicit imagery, and speeds up the implementation of required watermarking of AI-generated content from February 2, 2027, to December 2 of this year.
A new investigative report from 404 Media has exposed a sophisticated realtime deepfake tool that can run during live calls in WhatsApp, Zoom, and Microsoft Teams, among other platforms. Dubbed “Haotian AI”, the product is offered selectively through Chinese-language, Telegram-based networks that offer a range of illicit tools and services ranging from malware and money laundering services to “high-end” escorts. After successfully contacting a seller, 404 Media’s team was told they could rent the software for $1,998 a year, with payment to be made via the TRON cryptocurrency.
Apple is paying a price for how bad Siri is. The company has reached a $250 million settlement in a class action lawsuit accusing it of misleading iPhone buyers by advertising advanced Apple Intelligence and Siri features that were not available when the iPhone 16 launched. Apple’s marketing around the 2024 iPhone launch had promised a more personalized, AI-powered personal assistant on the device, only to delay the upgrade and admit that it would take “longer than we thought” to deliver the promised features. The settlement does not include an admission of wrongdoing, still requires court approval, and could pay eligible U.S. iPhone buyers roughly $25 to $95 per device, depending on the number of claims.
Finally, we can’t close out this week’s AI Update without a selective excerpt from the text messages between Sam Altman and Mira Murati that resurfaced during week two of the trial for Elon’s lawsuit. The texts revolve around the period in 2023 when Altman was briefly deposed from his role as CEO by the OpenAI board and was trying to reach out to his compatriot about what was going on behind the scenes. People have been having fun with them on social media:
–
By Alex Perala
