Technology

➚ MOVE FAST: Sweetheart deal. Wiz not only got Google to pay up for its second try at buying the cybersecurity firm, but it will also get a whopping $3.2 billion if the search giant fails to win regulatory approval from the Trump administration and others.

➘ BREAK THINGS: Deal-breaker. Samsung’s CEO said it will pursue acquisitions to prop up the company’s lagging growth, after falling behind rivals in the AI race. It’s a stark acknowledgment that the company’s strategies to catch up to competitors have failed.

Arizona’s Supreme Court has developed two digital avatars to explain court rulings, in one of the first cases of the justice system utilizing AI to inform the public, the Associated Press reported. The decision follows a protest last year around the court’s ruling to criminalize nearly all abortions — a verdict that the chief justice told the AP some residents misunderstood.

The court can create videos of Daniel and Victoria, the AI-powered avatars, in roughly 30 minutes — much quicker than producing a video of a human spokesperson. Powered by Silicon Valley-based Creatify, which uses AI to develop video advertisements, the avatars follow a script determined by a justice and the communications team rather than interpreting court decisions on their own.

It’s too early to determine whether the avatars have resonated with the public. The court has only posted a handful of videos that received limited engagement.

Video explanations shared on social media, however, can be more accessible to individuals than opinion documents riddled with legal jargon and the accompanying written releases often shared by higher courts. With many Americans believing the justice system is broken, clear and concise explanations may help courts regain the public’s trust.

Nvidia and Google DeepMind announced a handful of new collaborations this week, including one involving a real world Star Wars droid that shared the stage with Jensen Huang Tuesday.

One that got relatively little notice was a partnership on an AI watermarking technology developed by Google called SynthID. Announced back in 2023, SynthID uses some kind of AI model (the details are secret) to infuse images with an imperceptible watermark that can’t be removed by making simple changes to the file — a shortcoming in many watermarking technologies.

We’re more than two years into the ChatGPT era of generative AI and it still feels like we’re a long way from some kind of standard that makes it easy to detect AI-generated content.

When companies like Google and Nvidia agree on a piece of technology, it seems like the seeds of progress in the space.

I had an interesting conversation with Pushmeet Kohli, VP of research in DeepMind’s AI for Science division, about the partnership with Nvidia earlier this week and the similarities between this kind of technology and standards like HTTPS that helped make the web safe.

Kohli said it’s probably too early for something like that for AI detection. “If you had asked me this question four, five years back, whether this is even possible, I would have told you, I don’t know, because there’s genuine uncertainty about the hardness of the problem that we’re trying to solve,” he said.

What this might ultimately look like is a watermarking standard like SynthID that gets added by companies that serve AI models. For instance, Google Cloud could add SynthID to every image it produces, even on non-Google AI models. It’s not doing that today.

You could even see something happening at the chip level with Nvidia. But we’re probably a long way off from anything like that happening.

The percent of employees using AI at work who believe their companies’ AI rollouts in the last year have been successful, significantly lower than the 75% of C-suite executives who felt the same, according to research from enterprise AI company Writer. The discrepancy highlights that executives are often disconnected from how workers feel about the effectiveness of AI. Some 35% of employees who use AI at work are paying out of pocket for those tools, rather than using company-provided ones, the study showed.

World, the blockchain-enabled company with the famous face-scanning orb, is bringing its services to the video game world in a partnership with computer maker Razer, allowing users to verify their identity with World ID.

In case you need a refresher, World makes an eyeball-scanning orb that creates a unique identity for each person who uses it. In countries where it’s allowed, participants receive a cryptocurrency called Worldcoin. Little encrypted fragments created from each scan are scattered across a distributed ledger, making it very difficult to hack or reverse engineer someone’s identity.

In the age of AI, determining bots from humans is a growing challenge. World allows people to verify that they are a unique individual while not divulging any personally identifiable information.

Online gaming is a good testing ground for the tech because the space is filled with bots and bad behavior. Theoretically, if an anonymous user were harassing people in an online video game, they could have their World ID banned, preventing them from just creating another anonymous account and returning.

Razer has a single sign-on feature that allows people to log into multiple video games with one Razer login. Razer can now ensure that users who authenticate with World ID — even if they are anonymous and have multiple accounts — represent a single, actual human being.

For now, the Razer partnership is more of an experiment. The first video game to participate in the Razer-World collaboration is Tokyo Beast. Users don’t have to have a World ID to play, and having one only gets people special perks. Eventually, though, World and other methods of anonymous verification could become the norm — which could change the online gaming experience dramatically.

CIOs and IT leaders don’t just follow trends — they drive them. That’s why IT leaders at global organizations subscribe to CIO Upside. Get exclusive insights into the innovations and strategies shaping the future of tech leadership — subscribe for free today.

Elon Musk’s xAI is now part of the global AI Infrastructure Partnership, a $30 billion investment effort launched by Microsoft, BlackRock, and MGX last fall — a new development sure to sting OpenAI, once an inseparable partner to Microsoft. Musk is suing OpenAI, hoping to prevent it from becoming a fully formed company, and he’s been prodding it with hostile takeover offers.

This is all about chips. OpenAI was forced to go elsewhere when Microsoft was unable to procure enough Nvidia GPUs, while Microsoft has been irked by OpenAI’s disloyalty.

But the bigger takeaway is that some of the most serious players in AI are behind the infrastructure partnership. There is only one frontier foundation model company included, and it’s not Anthropic or OpenAI.

On stage during his keynote Tuesday, Nvidia CEO Jensen Huang bragged about the company’s role in building “Colossus,” the xAI data center in Memphis, the largest supercomputer built to date. Musk’s company is now hoping to grow that cluster ten-fold, to one million GPUs, likely supplied by Nvidia.

Colossus was built in 122 days. Nvidia is clearly on Musk’s side. If it doesn’t have the world’s best AI researchers today, it will surely have more success recruiting when AI researchers cannot resist greater compute power.

The biggest bottleneck to Musk’s supersonic speed is essentially energy, which the infrastructure partnership aims to rectify.

Two months into President Donald Trump’s second term, there’s already plenty of chatter about the odds that Vice President JD Vance will be his successor, Semafor’s Burgess Everett and Shelby Talcott report.

Vance is locking up some key support — but he’s also got a ton of time between now and a potential 2028 primary.

That’s both an opportunity and a challenge: If Vance can push through Trump’s agenda and help him protect GOP majorities in the midterms, he’ll be in prime position to win a Trump endorsement and take the MAGA mantle.