Hello and welcome to Eye on AI. In this edition….a dispatch from Davos…OpenAI ‘on track’ for device launch in 2026…Anthropic CEO on China chip sales…and is Claude Code Anthropic’s ChatGPT moment?

Hi. I’m in Davos, Switzerland, this week for the World Economic Forum. Tomorrow’s visit of U.S. President Donald Trump is dominating conversations here. But when people aren’t talking about Trump and his imposition of tariffs on European allies that oppose his attempt to wrest control of  Greenland from Denmark, they are talking a lot about AI.

The promenade in this ski town turns into a tech trade show floor at WEF time, with the logos of prominent software companies and consulting firms plastered to shopfronts and signage touting various AI products. But while last year’s Davos was dominated by hype around AI agents and overwrought hand-wringing that the debut of DeepSeek’s R1 model, which happened during 2025’s WEF, could mean the capital-intensive plans of the U.S. AI companies were for naught, this year’s AI discussions seem more sober and grounded.

The business leaders I’ve spoken to here at Davos are more focused than ever on how to drive business returns from their AI spending. The age of pilots and experimentation seems to be ending. So too is the era of imagining what AI can do. Many CEOs now realize that implementing AI at scale is not easy or cheap. Now there is much more attention on practical advice for using AI to drive enterprise-wide impact. (But there’s still a tinge of idealism here too as you’ll see.) Here’s a taste of some of the things I’ve heard in conversations so far:

CEOs take control of AI deployment

There’s a consensus that the bottom-up approaches—giving every employee access to ChatGPT or Microsoft Copilot, say—popular in many companies two years ago, in the initial days of the generative AI boom, are a thing of the past. Back then, CEOs assumed front line workers, closest to the business processes, would know how best to deploy AI to make them more efficient. This turned out to be wrong—or, perhaps more accurately, the gains from doing this tended to be hard to quantify and rarely added up to big changes in either the top or bottom line.

Instead, top-down, CEO-led initiatives aimed at transforming core business processes are now seen as essential for deriving ROI from AI. Jim Hagemann Snabe, the chairman of Siemens and former co-CEO at SAP, told a group of fellow executives at a breakfast discussion I moderated here in Davos today that CEOs need to be “dictators” in identifying where their businesses would deploy AI and pushing those initiatives forward. Similarly, both Christina Kosmowski, the CEO of IT and business data analytics company LogicMonitor, and Bastian Nominacher, the cofounder and co-CEO of process mining software company Celonis, told me that board and CEO sponsorship was an essential component to enterprise AI success.

Nominacher had a few other interesting lessons, including how, in research Celonis commissioned, establishing a center of excellence for figuring out how to optimize work processes with AI resulted in an 8x better return than for companies that failed to set up such a center. He also said that having data in the right place was essential to running process optimization successfully.

The race to become the orchestration layer for enterprise AI agents

There is clearly a race on among SaaS companies to become the new interface layer for AI agents that work in companies. Carl Eschenbach, Workday’s CEO, told me that he thinks his company is well-positioned to become “the front door to work” not only because it sits on key human resources and financial data, but because the company already handled onboarding, data access and permissioning, and performance management for human workers. Now it can do the same for AI agents.

But others are eyeing this prize too. Srini Tallapragada, Salesforce’s chief engineering and customer success officer, told me how his company is using “forward deployed engineers” at 120 of Salesforce’s largest customers to close the gap between customer pain points and product development, learning the best way to create agents for specific industry verticals and functions that it can then offer to Salesforce’s wider customer base. Judson Althof, Microsoft’s commercial CEO, said that his company’s Data Fabric and Agent 365 products were gaining traction among big companies that need an orchestration layer for AI agents and a unified way to access data stored in different systems and silos without having to migrate that data to a single platform. Snowflake CEO Sridhar Ramaswamy meanwhile thinks the deep expertise his company has in maintaining cloud-based data pools and controlling access to that data combined with newfound expertise in creating its own AI coding agents, make his company ideally suited to win the race to be the AI agent orchestrator. Ramaswamy told me his biggest fear is whether Snowflake can keep moving fast enough to realize this vision before OpenAI or Anthropic move down the stack—from AI agents into the data storage—potentially displacing Snowflake.

A couple more insights from Davos so far: while there is still a lot of fear about AI leading to widespread job displacement, it hasn’t shown up yet in economic data. In fact, Svenja Gudell, the chief economist at recruiting site Indeed, told me that while the tech sector has seen a huge decline in jobs since 2022, that trend predates the generative AI boom and is likely due to companies “right sizing” after the massive pandemic-era hiring boom rather than AI. And while many industries are not hiring much at the moment, Gudell says global macroeconomic and geopolitical uncertainty are to blame, not AI.

Finally, in a comment relevant to one of this week’s bigger AI news stories—that OpenAI is introducing ads to ChatGPT—Snabe, the Siemens chairman had an interesting answer to a question about how AI should be regulated. He said that rather than trying to regulate AI use cases—as the EU AI Act has done—governments should mandate more broadly that AI adhere to human values. And the one piece of regulation that would do more than anything to ensure this, he said, would be to ban AI business models based on advertising. Ad-based AI models will lead companies to optimize for user engagement with all of the negative consequences for mental health and democratic consensus that we’ve seen from social media, only far worse. 

With that, here’s more AI news.

Jeremy Kahn
jeremy.kahn@fortune.com
@jeremyakahn

Beatice Nolan wrote the news and sub-sections of Eye on AI.

FORTUNE ON AI

Wave of defections from former OpenAI CTO Mira Murati’s $12 billion startup Thinking Machines shows cutthroat struggle for AI talent–by Jeremy Kahn and Sharon Goldman

ChatGPT tests ads as a new era of AI begins—by Sharon Goldman

A filmmaker deepfaked Sam Altman for his movie about AI. Then things got personal—by Beatrice Nolan

PwC’s global chairman says most leaders have forgotten ‘the basics’ as 56% are still getting ‘nothing’ out of AI adoption–by Diane Brady and Nick Lichtenberg

AI IN THE NEWS

OpenAI ‘on track’ for device launch in 2026. OpenAI is hoping to unveil its first-ever consumer device in the second half of 2026, according to chief global affairs officer Chris Lehane, who spoke at Axios House Davos on Monday. It’s the clearest timeline yet for the secretive project that CEO Sam Altman has been teasing since partnering up with former Apple design chief Jony Ive last year. Details remain tightly guarded, but reports suggest OpenAI is experimenting with small, screenless devices—possibly wearable—that rely on AI-driven interaction rather than traditional apps or displays. While Lehane didn’t confirm an exact commercial launch date, he said the "most likely" window is later this year, depending on how "things advance."

Selling AI chips to China is like 'selling nuclear weapons to North Korea,' Anthropic CEO says. Dario Amodei has criticized the Trump administration's move to ease restrictions on advanced AI chip exports to China, calling it "a big mistake" with "incredible national security implications" in an interview with Bloomberg at the World Economic Forum. In a panel discussion later in the day, Amodei doubled down on his stance–saying that restricting the sale of advanced chips to China was the single best thing the U.S. could do to secure its lead in the global AI race. He has previously warned about dystopian scenarios if authoritarian regimes gain access to advanced AI capabilities.

Musk $134 billion lawsuit hits a nerve at OpenAI. Elon Musk is seeking between $79 billion and $134 billion in damages from OpenAI and Microsoft, according to a report from Bloomberg. In a new court filing, Musk’s lawyers argue he is entitled to a share of OpenAI’s estimated $500 billion valuation after being “defrauded” of the $38 million he donated as a co-founder. OpenAI and Microsoft have dismissed the claims as baseless and “unserious,” accusing Musk of harassment ahead of a jury trial scheduled for April. Meanwhile, some of the discovery released by Musk seems to have struck a nerve with the lab’s leadership. On Friday, OpenAI released a detailed blog post accusing Musk of misrepresenting internal documents and omitting key context showing he supported a for-profit structure as early as 2017.

Anthropic gains one of OpenAI’s key safety researchers. Anthropic has hired Andrea Vallone, a senior safety researcher who led OpenAI’s work on how chatbots should respond to signs of emotional over-reliance and mental health distress. Vallone spent three years at OpenAI building its “model policy” research team and helping shape the deployment and training of GPT-4, reasoning models, and GPT-5. Vallone will join Anthropic’s alignment team under Jan Leike, the former OpenAI safety lead who left in 2024, citing concerns that safety was being overshadowed by product pressures. The move comes amid heightened scrutiny of how AI systems handle vulnerable users, following lawsuits and congressional attention tied to chatbot interactions and mental health harms. More from The Verge here.

EYE ON AI RESEARCH

Researchers say ChatGPT has a "silicon gaze" that amplifies global inequalities. A new study from the Oxford Internet Institute and the University of Kentucky analyzed over 20 million ChatGPT queries and found the AI systematically favors wealthier, Western regions, rating them as "smarter" and "more innovative" than poorer countries in the Global South. The researchers coined the term "silicon gaze" to describe how AI systems view the world through the lens of biased training data, mirroring historical power imbalances rather than providing objective answers. They argue these biases aren't errors to be corrected, but structural features of AI systems that learn from data shaped by centuries of uneven information production, privileging places with extensive English-language coverage and strong digital visibility. The team has created a website–inequalities.ai–where people can explore how ChatGPT ranks their own neighborhood, city, or country across different lifestyle factors.

AI CALENDAR

Jan. 19-23: World Economic Forum, Davos, Switzerland.

Jan. 20-27: AAAI Conference on Artificial Intelligence, Singapore.

Feb. 10-11: AI Action Summit, New Delhi, India.

March 2-5: Mobile World Congress, Barcelona, Spain.

March 16-19: Nvidia GTC, San Jose, Calif.

BRAIN FOOD

Is Claude Code Anthropic's ChatGPT moment? Anthropic has started the year with a viral moment most labs dream of. Despite Claude Code's technical interface, the product has captured attention beyond the developer pool, with users building personal websites, analyzing health data, managing emails, and even monitoring tomato plants—all without writing a line of actual code. After several users pointed out that the product was much more of a general-use agent than the marketing and name suggested, the company launched Cowork—a more user-friendly version with a graphical interface built for non-developers.

Both Claude Code and Cowork's ability to autonomously access, manipulate, and analyze files on a user's computer has given many people a first taste of an AI agent that can actually take actions on their behalf, rather than just provide advice. Anthropic also saw a traffic lift as a result. Claude's total web audience has more than doubled from December 2024, and its daily unique visitors on desktop are up 12% globally year-to-date compared with last month, according to data from market intelligence companies Similarweb and Sensor Tower published by The Wall Street Journal. But while some have hailed the products as the first step to getting a true AI personal assistant, the launch has also sparked concerns about job displacement and appears to put pressure on a few dozen startups that have built similar file management and automation tools.

FORTUNE AIQ: THE YEAR IN AI—AND WHAT'S AHEAD

Businesses took big steps forward on the AI journey in 2025, from hiring Chief AI Officers to experimenting with AI agents. The lessons learned—both good and bad–combined with the technology's latest innovations will make 2026 another decisive year. Explore all of Fortune AIQ, and read the latest playbook below: 

–The 3 trends that dominated companies’ AI rollouts in 2025.

–2025 was the year of agentic AI. How did we do?

–AI coding tools exploded in 2025. The first security exploits show what could go wrong.

–The big AI New Year’s resolution for businesses in 2026: ROI.

–Businesses face a confusing patchwork of AI policy and rules. Is clarity on the horizon?