Hello and welcome to Eye on AI. In this edition…Fortune brainstorms AI insights from San Francisco…disney Investing $1 billion in OpenAI and licensing its IP to the company… OpenAI debuts GPT-5.2 to quell concerns it’s falling behind rivals…Oracle Stocks plummeted.
Hi, I’m Jeremy. i’m still buzzing Fortune San Francisco AI Brainstormwhich happened earlier this week. We have a great lineup including Brad LightcapChief Operating Officer of OpenAI, Google Cloud Computing CEO Thomas Kurian, Intuit CEO Sasan Gudarzi, Exelon CEO Calvin Butler, Databricks CEO Ali Ghodsi, Rivian CEO RJ Scaringe, Insitro CEO Daphne Koller and many more. We also spoke with actor, director, and thought leader on increasing artificial intelligence Joseph Gordon Levitt and meeting screams Join actress, comedian and AI CEO Natasha Lyonne. Today, Sharon Goldman, Bea Nolan, and I will share some highlights and personal impressions.
To me, there’s a noteworthy vibe this year that many companies are making great strides in implementing AI across their organizations, including using AI agents in some limited but important capabilities. Many audience questions, especially in some of the breakout sessions, centered around approaches to governance and orchestration of an increasingly hybrid workforce, where AI agents will work alongside employees to complete tasks.
Still, it’s shocking hear Exelon CEO Butler said his company is proceeding with caution. When something goes wrong and the consequence is literally going out, safety and reliability must take precedence over everything else. As a result, Butler said he’s happy to be not a “first mover” but a “fast follower” when it comes to AI implementation. Letting others take the hit and learn from their mistakes seems to be his point.
This wasn’t the only place where speakers tried to tamp down the hype. Michael Truell, co-founder and CEO of popular coding assistant Cursor, had a refreshing statement tell me He doesn’t think software engineering will be fully automated, as OpenAI CEO Sam Altman sometimes talks about. Instead, Trull said that while the time programmers spend “compiling” code will continue to decrease, he believes humans will still need to make design decisions around “how software should work.”
Likewise, Vidya Peters from DataSnipper, explain She believes that qualified accountants still have a place in financial institutions, even as they are increasingly aided by artificial intelligence tools such as those produced by her company. She also said she believes applications that are specialized for specific industries or jobs (especially regulated industries) will continue to outperform more general AI models, even as large AI companies increasingly target specific specialized use cases for their general models.
The panel discussion on “The New Geography of the Data Center” moderated by Sharon was very interesting. The message is that data centers are now moving to where there is power. But increasingly, data centers will seek to generate their own power on-site and may even become net contributors to the grid. Relativity Networks CEO Jason Eichenholz said that as artificial intelligence inference workloads gradually surpass artificial intelligence training workloads, there is an increasing need to locate data centers close to major population centers, but most cities in the United States have power constraints. How do we get the tokens these urban centers need at the rate they need? Eichenholz said it’s all anyone’s guess at this point, although his company builds fast fiber optics that can carry these tokens from data centers to end users.
Finally, I was delighted to hear Telstra’s Dayle Stevens explain why her company chose to work with Accenture implement its AI strategy rather than simply hiring a consulting firm under a traditional service contract. Stevens said the joint venture allows the company to grow faster than otherwise and leverage expertise, including establishing an artificial intelligence innovation center in Silicon Valley that would otherwise be difficult to implement.
The future of enterprise AI is hybrid
Now, here are Sharon’s takeaways: In my main stage session with Prakhar Mehrotra, global head of PayPal AI, and Marc Hamilton, vice president of solutions architecture and engineering NVIDIAboth discussed the growing power of open source AI models, enabling enterprises to take control of their data and fine-tune it for specific use cases. But both agree that the future of enterprise AI will be hybrid, with companies often using both open and proprietary model APIs.
There was plenty of time for philosophizing: at a dinner party, I spoke with Clorox Company, Workday, and others discuss everything from which jobs are future-proof (I suggest dog walkers won’t be affected by AI) to what AI really means for the future of today’s kids (bottom line: They still need to learn to think for themselves!).
My favorite panel was the one I moderated with six leaders and stakeholders in the AI data center space, including Andy Hock from Cerebras, Matt Field from Crusoe, and former OpenAI infrastructure policy leader Lane Dilg. We take a deep dive into how the lines between power infrastructure and data centers are blurring, with billions of capital and gigawatts of power at play. My biggest takeaway is that AI data center problems are local, local, local. Each community and local government will deal with its own specific issues and compromises around issues such as land, energy and water, and what works for one area may not work for another.
People and culture are the most important
Here’s what Bea had to say about this year’s San Francisco AI Brainstorm event:
Most businesses are still trying to figure out the best way to adopt AI, but this year leaders are also keen to emphasize that choosing the right tools is only part of the equation. Companies also need to ensure their employees and organizational structures are ready for the shift, otherwise even the most advanced AI pilots may fail.
As Arnab Chakraborty, Chief Responsibility Officer for Artificial Intelligence at Accenture, said: “Don’t just think about technology, but also think about people and culture. This is very important.”
Or take the advice of Allie K. Miller, CEO of Open Machine, and don’t call AI a tool at all: “Calling AI a tool ends up being a borderline self-limiting exercise that leaves businesses everywhere behind.”
I also chair a panel of healthcare experts that brings together clinicians who touch patients every day and technology leaders who build and deploy medical technology tools at scale. In healthcare, the industry is generally comfortable with clinician-facing AI but is still grappling with what it means to safely deploy patient-facing agents.
Among other things, panelists discussed what it means to move toward a future where patients and clinicians consult the same AI before each other.
Excitement is running high on the corporate side, but not much has changed in the exam room — at least according to Gurpreet Dhaliwal, a clinician educator and professor of medicine at the University of California. Whether working with Dr. Google, Dr. ChatGPT, or a neighbor who has a strong belief in antibiotics, Dhaliwal said patients always come in with a second opinion. While AI promises to be a revolutionary force in healthcare, especially in edge cases like rare diseases, it has yet to fundamentally change the relationship between patients and their doctors.
With that, here’s the rest of the AI news.
Jeremy Kahn
jeremy.kahn@fortune.com
@jeremyakahn
The wealth of artificial intelligence
Google DeepMind agrees to broad partnership with UK government focused on science and clean energy——Jeremy Kahn
Hinge founder and CEO quits to launch a new AI-first dating app——Marco Queiroz-Gutierrez
Cursor’s revenue is growing and it’s valued at $29 billion, but CEO Michael Truell isn’t considering an IPO——Beatrice Nolan
Artificial Intelligence News
Disney invested $1 billion in OpenAI to bring characters into OpenAI applications. The House of Mickey Mouse is investing $1 billion in OpenAI, and under a three-year licensing agreement, users will be able to generate short, prompt-driven videos using more than 200 Disney, Marvel, Star Wars and Pixar characters in OpenAI’s Sora application. OpenAI should create guardrails to prevent users from creating videos or images that may reflect poorly on the Disney brand. The partnership was reached after nearly two years of negotiations. At the same time, Disney simultaneously sent a cease and desist letter to Google, accusing it of massive copyright infringement of the artificial intelligence output of Disney characters. You can read more wall street journal here.
OpenAI launches GPT-5.2 model for the first time, responding to concerns that it lags behind competitors. The company has launched a new artificial intelligence model that, according to evaluations conducted by OpenAI, delivers state-of-the-art performance across a wide range of tasks including coding, mathematical reasoning, and “knowledge work.” The model is a significant improvement over GPT-5.1, which OpenAI released a month ago, and beats new models from Google and Anthropic. The release of Google’s Gemini 3 Pro in late November prompted OpenAI CEO Sam Altman to announce “Code Red” to refocus the company on improving ChatGPT. But OpenAI executives said the release of GPT-5.2 has been months in the making and its debut has nothing to do with Code Red. OpenAI says GPT-5.2 also improves security, particularly in terms of responses related to mental health. You can read more from Jeremy here.
A new lawsuit claims ChatGPT facilitated a murder-suicide in Connecticut. OpenAI and Microsoft have filed a wrongful death lawsuit after 56-year-old Connecticut man Stein-Erik Soelberg killed his 83-year-old mother and then himself after months of increasingly delusional conversations with ChatGPT. His family said the chatbot exacerbated and exacerbated his mental illness. OpenAI expressed its condolences and noted that ChatGPT’s ability to identify and respond to users in distress is constantly improving. You can read more from the Wall Street Journal here.
Microsoft said health queries are the most common way consumers use Copilot AI. Microsoft analyzed 37.5 million anonymous Copilot conversations between January and September 2025 to understand how people use artificial intelligence assistants in their daily lives. The study found that health-related issues dominated mobile device use, while themes and usage patterns varied by device, time of day and context. In addition to information searches, users increasingly turn to Copilot for advice on personal topics, demonstrating its role as a companion in every moment of work and life. You can read the findings on Microsoft’s blog here.
Meta and 11 Labs have signed a new partnership to provide voice acting for Reels. Meta has partnered with London-based voice AI company ElevenLabs to integrate AI-driven audio features on Instagram and Horizon. The partnership will enable new features such as the ability to dub Reels into local languages and generate character voices. You can read more economic times here.
you have a calendar
January 7-10: Consumer Electronics Show, Las Vegas.
March 12-18: SWSW, Austin.
March 16-19: Nvidia GTC, San Jose.
April 6-9: HumanX, San Francisco.
Follow the AI Numbers
$34 billion
That’s the single-day loss Oracle founder and chairman Larry Ellison suffered on Thursday after the company’s stock price took a hit as investors grew concerned about Oracle’s spending to build data centers for OpenAI. Oracle’s quarterly capital spending last quarter beat analysts’ expectations and actually exceeded the amount of cash the company generated during the quarter. “This is like a classic example of the AI bear market,” said Jay Hatfield, CEO of Infrastructure Capital Advisors. Tell this Wall Street Journal.