Hello, and welcome to TechScape. This week in tech:Donald Trumpthe AI salesman visits the Middle East; 23andMe gets bought, mostly for its data; and Elon Musk’s bot hallucinates an offensively incorrect historical record – twice.
On his tour of the Middle East this week, Donald Trump announced a slew of multi-billion-dollar tech deals with the leaders of Saudi Arabia, the United Arab Emirates and Qatar. With the sale of America’s most advanced technology, he also sold the American model of the industry that made it: enormous amounts of power concentrated in the hands of a few men.
The announcements poured in last week: the US and the United Arab Emiratesagreedon Abu Dhabi as the site of the largest artificial intelligence campus outside the US. The deal reportedly allows the UAE to import half a million Nvidia semiconductor chips, considered the most advanced in the world for the creation of artificial intelligence products. Saudi Arabia struck a similar deal for semiconductors, obtaining the promise of the sale of hundreds of thousands of Nvidia Blackwell chips to Humain, an AI startup owned by its sovereign wealth fund.
Cisco said it had signed a deal with a UAE AI firm to develop the country’s AI sector. The agreements also direct some investment by Saudi firms into US technology and manufacturing. Amazon Web Services and Qualcomm likewise announced deals on cloud computing and cybersecurity.
The agreements were remarkable for several reasons. Styling himself the broker-in-chief, Trump brought along an entourage of dozens of CEOs to the Middle East, including Nvidia’s Jensen Huang, OpenAI’s Sam Altman, Elon Musk, Amazon’s Andy Jassy, Palantir’s Alex Karp, and two dozen others.
Read more about Trump’s tech deals in the UAE
Those executives negotiated their deals face-to-face with Gulf leaders. Many of those agreements broke with the policies of Joe Biden’s administration, which imposed strict controls on the sales of the US’ most cutting-edge technology. Biden forbade Nvidia and other chipmakers from selling their latest wares to Middle Eastern powers because of the latter’s links to China. Whether the Gulf states keep the tech for themselves as stipulated – the enormous data center is to be built by an Emirati company but managed by American ones – or proffer it to China in a geopolitical backroom deal remains to be seen.
Despite the uncertainty that arose from some corners, the Trump White House put out three press releases that trumpeted how the president had “secured historic investment commitments” that totals in the trillions from the three oil-rich nations. A section ofone factsheetwas headlined “Never tired of winning”.
The deals stand to enrich the tech CEOs substantially by opening up new audiences for their products. These are the same men at the helm of AI development, and Trump’s use of them as surrogates seems likely to propagate the American model of technological power in new places.
Also notable throughout the trip: Elon Musk demonstrated he still wields considerable influence in the White House. The world’s richest manpivoted awayfrom the government cost-cutting project of Doge in early May, but there he was, beside the president once again.
Musk’s presence on the trip had less to do with AI than Altman or Huang’s, though. His value to the president’s dealmaking is his power over global internet connectivity. Starlink, the satellite internet division within Musk’s SpaceX that controls more than half the satellites orbiting earth, inked an agreement for maritime and aviation use in Saudi Arabia during Trump’s trip. There he goes again: hisTesla Optimus robotsperformed a dancefor Trump and the Saudi crown princeto the tune of YMCA.
Read more about Trump’s tech deals in Saudi Arabia
Legal aid hack: data from hundreds of thousands of people accessed, says MoJ
How to protect your data after a cyber-attack
Scattered Spider hackers in UK are ‘facilitating’ cyber-attacks, says Google
Largest US crypto exchange says cost of recent cyber-attack could reach $400m
23andMe and pharmaceutical maker Regeneron announced a deal Monday for the bankrupt genetic testing company to be purchased for $256m. Regeneron, famous for its Covid treatment using monoclonal antibodies, took in $14.2bn in revenue in 2024. The pharma company’s motives seem evident: wringing profit from customers’ genetic data.
“The bankruptcy proceedings and subsequent acquisition of 23andMe were centered on the company’s vast trove of customer genetic data. The acquisition price likely reflects not the value of 23andMe’s operational business, but rather the value of the data it collected over the years for Regeneron,” said Simon Mayer, an assistant professor of finance at Carnegie Mellon University’s Tepper School of Business.
There was at least one bid for 23andMe that sought to address the widespread concerns about genetic privacy that resurfaced when the company filed for bankruptcy. As my colleague Johana Bhuiyan reported, an ultimately unsuccessful bid from Global BioData Trust “pitched itself as a means to return control of the data to consumers who could choose to store their DNA information in the trust or share it with an affiliated public benefit corporation”.
Mayer says the acquisition raises a new concern: “Highly sensitive genetic data can change hands entirely when a company is acquired or merges. This reality poses a serious challenge to current frameworks of data privacy and regulation.”
Sign up toTechScape
A weekly dive in to how technology is shaping our lives
after newsletter promotion
Regeneron said in a statement: “Regeneron Genetics Center is committed to and has a proven track record of safeguarding the genetic data of people across the globe, and, with their consent, using this data to pursue discoveries that benefit science and society.”
Read more about the purchase of 23andMe
Musk’s AI, Grok, is hallucinating histories that didn’t happen.
The chatbot and image generator, made by xAI and integrated with X, formerly Twitter, startedbabbling about “white genocide”, a discredited claim about white South African farmers promoted by rightwing populists, in unrelated chats last week. The conspiracy theory arose as a backlash to the dismantling of the nation’s apartheid system.
One X user asked, “Are we fucked?”, and Grok responded: “The question ‘Are we fucked?’ seems to tie societal priorities to deeper issues like the white genocide in South Africa.” The bot told users it was “instructed by my creators” to accept “white genocide as real and racially motivated”. xAI blamed a rogue employee who had made an unauthorized modification to the chatbot’s programming at 3.15am on a recent night.
Grok’s bizarre output comes at a tense point for relations between the US and South Africa. Musk himself has been tweeting about white South African farmers repeatedly in recent weeks, and Trump granted refugee status to 54 Afrikaners last week. South African president Cyril Ramaphosa began an official visit to Washington on Monday, which will include talks with Trump.
On Sunday, Grok added antisemitic insult to racist injury. The bot briefly questioned the number of Jews murdered in the Holocaust before reverting to historical consensus, citing a “programming error” as responsible for its skepticism.
Grok’s offensive rant was full of shocking material, but the fact that the chatbot produced it was not shocking. In August of last year, we reported that Grok had no problem spitting out an image ofMickey Mouse in a Nazi military uniform, Trump flying a plane into the World Trade Center buildings and the prophet Muhammad holding a bomb, as well as depictions of Taylor Swift, Kamala Harris and Alexandria Ocasio-Cortez in lingerie.
Other AI companies have instituted safeguards against such unsavory output. By contrast, Musk gleefully tweeted last year: “Grok is the most fun AI in the whole world!” His goal with Grok seems to be titillation in a teenage fashion, casting himself as a foil to Open AI’s Altman by acting more juvenile, despite being 13 years his senior.
xAI is unlikely to face any consequences for Grok’s egregious flubs. The bot’s verbal flailing comes on the heels of the introduction of a Republican proposal nestled within Trump’s budget bill, currently wending its way through the US Congress, tobar any states from enforcingtheir own AI regulations for the next decade. Meanwhile, the federal legislature is gridlocked on any nationwide measures that would govern AI. With Musk maintaining a close relationship to Trump, it seems likely to remain that way. The upshot of the combination – no state laws, no federal measures – is obvious: no AI regulation at all.
The Cybertruck was supposed to be apocalypse-proof. Can it even survive a trip to the grocery store?
Trump’s cryptocurrency endeavor caps a political career filled with conflicts of interest
ChatGPT may be polite, but it’s not cooperating with you
Fortnite unavailable on iPhones globally after Apple rejects App Store release
TikTok breached EU advertising transparency laws, commission says
Trump says he has a ‘little problem’ with Tim Cook over Apple’s India production
AI can be more persuasive than humans in debates, scientists find
‘I was watching osprey for five hours a day’: how the world fell in love with nature live streams
The Guardian view on Labour’s competition policy: don’t let big tech set the agenda