AI news and live updates

Factual accuracy and political bias remain significant issues for AI consumer-focused outputs.

Forum AI Founder and CEO Campbell Brown said recent research by her firm suggests that 40% of all outputs had at least one major mistake or got one claim wrong.

“That’s not great,” she told CNBC’s Scott Wapner.

That is being amplified in the lead-up to the midterm elections, when people turn to chatbots or AI-powered search engines to ask where their polling places are, when they need to register, or what topics are on the ballot.

“You get those questions wrong, and you are taking away someone’s right to vote,” she said.

On top of that, Brown said that Forum AI’s research found instances of Anthropic’s Claude “pulling from the Chinese Communist Party website to answer questions about a U.S. election that had nothing to do with China.

“This, to me, is low-hanging fruit,” she said.

Wapner asked Brown whether AI labs are currently incentivized to correct that accuracy, or whether the focus on enterprise applications of AI is taking precedence.

“It’s been a low priority,” she said. “They’re making so much money from coding … this is a room full of CTOs, you know what they’re doing, and that’s where the emphasis has been. The consumer use cases have been less important.”

A TEC member asked Brown if she saw a future where the consumer becomes a priority for AI outputs, especially given the heavy adoption split so far between enterprise and consumer revenue generation.

“I don’t think the consumer is going to be a priority for a while; they’re just making too much money on the enterprise side, and they really don’t talk about political evals or election evals until we get public research and it gets picked up by the press,” Brown said. “It’s just not a priority and won’t be for a while until something happens.”

Brown pointed to how AI labs have reacted amid concerns around mental health, which she said has been “a wake-up call for them.”

“They sort of unleashed this and didn’t realize what a big use case that was going to be, especially with teenagers, and then you have these horrible incidents happening and litigation flooding into the labs,” she said. “So now they’re bringing in clinicians and people who actually know what they’re talking about … it goes back to the importance of having real expertise in the process.”

Pointing to the lights above the stage at the 2026 CNBC Technology Executive Council Summit, Publicis Sapient CEO Nigel Vaz told CNBC’s Jon Fortt that one of the issues that is plaguing companies when it comes to AI use is thinking about usage like lighting.

“So many conversations around tokens are like if I said to you if you could pay me for three watts,” Vaz said. “The question you want to know is will that turn on that light bulb or not, because what I care about is paying you for a unit of work.”

Vaz took the lighting metaphor a step further when Fortt asked him how AI ROI measurement should work then.

“Most employees want to see that they’re creating value through the work that they’re generating. So effectively deploying AI as like a dimmer switch to make everybody just a little brighter,” he said. “We are already past that phase.”

Vaz said AI usage needs to be grounded in outcomes that workers can understand, “and then contextualized about how it’s going to change their lives in that context.”

“The other way around, which is I think where a lot of AI projects started was, ‘Let’s deploy AI more broadly across the enterprise. Let’s give lots of people access. Let’s run a lot of experiments,’” he said. “I think you get value, but you make everybody just a little bit brighter. You may not meaningfully move the needle on a specific outcome.”

— Ian Thomas

AI is creating new, fundamentally different operating models for businesses. But often AI engineers are not understanding the business in which they are operating, creating multiple issues when tracking ROI and success of the technology.

Nigel Vaz, Publicis Sapient CEO, told CNBC’s Jon Fortt that he was at a breakfast before the CNBC Technology Executive Council Summit where he met with a group of clients that his firm works with.

The CIO of an investment bank that Publicis Sapient works with told Vaz that the bank recently brought in some engineers from Anthropic.

“The client said, ‘I have 50,000 developers, and they gave me six incredible AI engineers from Anthropic. But they don’t understand investment banking. They don’t understand the bespoke coding language we use in the bank,’” Vaz said.

The client was highlighting something, Vaz said, that he continuously hears which is that people need to think about AI beyond the context of the technology, “because a technology or a tool given to an enterprise is just one part of the problem.”

“You need some people who understand the context of the business, not in terms of how it operates today, but who have a perspective on how it’s going to operate tomorrow because they’ve got to get somebody from here to there,” he said.

— Colleen McKown 

In a Wall Street Journal op-ed published in June, Forum AI Founder and CEO Campbell Brown argued that as the world becomes more reliant on AI, there needs to be a more critical lens applied to AI’s output.

“Already it is so good its mistakes don’t look like mistakes,” Brown wrote.

Speaking with CNBC’s Scott Wapner at the CNBC Technology Executive Council Summit in New York City on Wednesday, Brown said the issue remains that “when you read an answer generated by AI, it’s presented with such confidence and clarity that it is hard to question it and identify where the mistakes may be.”

“I think that’s what’s particularly dangerous,” she said.

Brown, the former head of global media partnerships at Meta, said this issue is increasingly worrying as people turn to AI for political or election information, and that speaks to how the information is not only presented but where it is sourced.

“There’s a real critical piece to this that has been somewhat frustrating for me: the labs are not engaging as much as I think they should be with real subject matter expertise,” she said. “Human expertise in domains is essential to getting this right.”

— Ian Thomas

Yesterday, CNBC’s Workforce Executive Council held its annual Summit, and the discussion of how human resources professionals help workers adapt to this new AI age was at the top of the agenda.

Walmart, the biggest private employer in the world, asked its 2.1 million associates in June that exact question for the first time, trying to get a sense of how one of the most diverse workforces is experiencing AI.

What it found, based on more than 1.7 million survey responses, is that while workers have different levels of familiarity and comfort with AI, training and experience with the technology matters.

Seventy-seven percent of Walmart’s office-based associates said are excited about AI’s potential to improve their work, Walmart Chief People Officer for Associate Experience and Digital Maren Waggoner told CNBC’s Leslie Picker at the WEC Summit.

And as for Walmart’s U.S. frontline associates, 84% said they are confident using digital tools, a reflection of the AI-powered tools that many of them are now using in their roles, she said.

“We’ve given them certificates, we’ve given them training programs and tools, and so they’re using it in their day-to-day,” Waggoner said.

Jill Bruning, chief technology officer of Amentum and a member of the CNBC Technology Executive Council, said at the WEC Summit that strategy should be evolved to be humans in the lead.

“There’s an AI curve, and we see it. It’s going up and it’s going fast and has a lot of attributes to it; it’s quicker than anything we’ve seen before,” Bruning told CNBC’s Susan Caminiti. “So when you think about people on that curve … that gap is getting bigger every day.”

Bruning said she doesn’t know if leadership is adapting fast enough to the technology, and that is creating additional challenges with upskilling as well as what tasks are being delegated to AI.

“So your delegation skills are different, and then how do we tell the agents what to do? How do we define that role? She said. “That’s human over the loop, not in the loop. We’re still in charge.”

For more from the CNBC WEC Summit, you can see the full live blog here.

— Ian Thomas

Here are some of the top tech-focused headlines from CNBC this morning:

Google alters its cloud playbook, enlisting partners to sell more AI chips. Google started working on in-house silicon in 2013, but didn’t begin selling its tensor processing units (TPUs) to outside customers until the second quarter of this year. Now the company is tapping Oracle’s cloud as another avenue for selling TPUs as it tries to bolster competition with Nvidia in the AI chip market.
Anthropic’s cash pile towers over nearly every U.S. company. Anthropic had $112.4 billion in cash and cash equivalents and $19.3 billion in short-term investments as of June 30. That exceeded every non-bank U.S. company in CNBC and S&P Capital IQ data except Berkshire Hathaway and Alphabet.
Amazon’s anti-tokenmaxxing memo says engineers should be pickier about using AI. Amazon discouraged teams from relying solely on token usage or code generation as measurements of AI adoption, calling them “deeply flawed” metrics in an internal memo. The memo revises how the company defines the role of software development engineers and managers in the age of AI.

— Ian Thomas

This year’s TEC Summit is focusing on what is coming next in AI, whether that’s how companies are adapting their strategies based on the advancement in the technology to how they’re controlling the costs related to that build out.

Discussions will also center on AI regulation, agents, how workers are using the technology and where this acceleration of AI is taking society.

Speakers include:

Campbell Brown, Forum AI Founder & CEO: Brown will be discussing AI regulation, independent evaluation, what that looks like and how companies can use AI that can be trusted.
Nigel Vaz, Publicis Sapient CEO: Vaz is at the center of how companies are rethinking their business models amid the AI age, and why that’s harder than learning how to use the technology itself.
Calix CEO Michael Weening and CPO Shane Eleniak: Members of the team at Calix will pull back the curtain on how strategic AI decisions are made and implemented across the organization.
Keia Cole, Brooklyn Sports & Entertainment Chief Digital Officer: Technology is also transforming the sports industry, and Cole will discuss how AI is being used at the Barclays Center and within the Brooklyn Nets and New York Liberty teams to make sports fans more connected and in touch.
Jason Gowans, Levi Strauss Chief Digital & Technology Officer: Organizations are aiming to empower workers and customers with AI tools. Gowans will be sharing his approach to embedding AI into the everyday shopping experience.
Jeetu Patel, Cisco President & Chief Product Officer: Rogue AI agents show no signs of slowing down, but Patel said companies need to think about them like they are teenagers. He’ll discuss ways companies can safely harness the power of agents and what it means to parent them.  

— Ian Thomas