Agents, OpenAI, deepfakes, and the strange reality of the AI boom: An interview with Oren Etzioni
[Editor’s Note: Agents of Transformation is an independent GeekWire series and March 24, 2026 event, underwritten by Accenture, exploring the people, companies, and ideas behind AI agents.]
Oren Etzioni got so frustrated with flipping between browser windows and following the step-by-step instructions for ChatGPT that he ended up asking: Do you work for me, or do I work for you?
ChatGPT’s response at the time: no, it couldn’t do the job for him. Etzioni, a computer scientist who has been building AI systems since the late 1980s, says bridging that gap between verbal AI and actionable AI is what defines this moment in the evolution of technology.
But as AI agents move from concept to reality, Etzioni says the “rough edge” of performance remains a stubborn problem: give an agent one request and it saves an hour and a half of work, then give it something almost identical and it produces garbage.
“We haven’t gained artificial credibility,” he said. “Those are still ways out.”
Etzioni spoke to GeekWire at an event hosted by Accenture in Bellevue, Wash., last week, with an audience that included Microsoft leaders. The University of Washington professor is the founder of the first AI agent Vercept, the founder and technical director of the AI2 Incubator, a business partner at Madrona, and the former CEO of the founder of the Allen Institute for AI.
In the evening, Etzioni asked questions about the emergence of the world of AI agents, platform competition between large technology companies, China’s expansion in AI research, and the emerging threat of democratic deepfakes. He also offered some sharp words for OpenAI and advice for leaders navigating AI discovery.
For agents: Etzioni said what works now is to send smaller, more streamlined workflows — the kind of tasks that used to require switching between apps and following instructions manually.
Vercept, for example, allows an agent to see what’s on your screen, find buttons, read text, and perform tasks directly, rather than relying on what it calls a “bad infrastructure” of APIs and web scraping that breaks whenever something changes.
The bigger picture is messier. Etzioni described Moltbook – a bot-only social network that attracted 1.6 million AI agents over the weekend – as overcrowded in its current state, but a signal of what’s to come: a future where software agents collaborate at scale.
He was blunt about the risks: Moltbook is a “security nightmare,” with agents running on users’ machines, accessing private information, and reading text sent outside of anyone’s control, making them vulnerable to instant injection attacks.
Etzioni pushed back on the current surprising intelligence, disagreeing with Microsoft AI CEO Mustafa Suleyman’s claim that we are witnessing the birth of a new digital genre.
“These are still tools,” he said. “Power tools, but tools still work in our place.”
In the speaker competition: Asked how he sees the race between Microsoft, Google, Amazon, OpenAI, and Anthropic, he said he would short OpenAI stock, if he did that.
“They’re running around like a thousand chickens with their heads cut off,” he said, questioning whether the company has a viable business model outside of its flagship chatbot. “Sure, they print money on ChatGPT, but that’s not their business.”
He’s advanced to Google, which he describes as having the virtue of vertical integration, from chips to data to models to talent. “They’re starting on the back foot,” he said, “but Google is ready to — I think it’s a technical phrase — kick their ass.”
Anthropic and OpenAI rush to IPOs as they burn cash. Once they go public, Etzioni noted, quarterly results will reveal who really wins.
In China: Etzioni said the belief that the country’s AI work is out of it is no longer true.
He cited research done by his team at the Allen Institute for AI, which tracks academic papers at top AI conferences, showing Chinese papers are increasing not just in volume but in quality. He said that trend has played out in open source models and technological innovation.
“I’m actually a China hawk – I’m very concerned about China’s role in the world,” he said. “But the solution is not to underestimate it, because that would be a mistake.”
For deepfakes: Etzioni spent more than a year running TrueMedia.org, a nonprofit he founded to build tools for news outlets and fact-checkers to detect deepfakes ahead of the 2024 election. The good news, he said, is that deepfakes didn’t significantly change the election results. The bad news is that the technology has become much cheaper and easier to use.
Looking ahead, he worries about “denying an attack on democracy” — not one viral fake but thousands of AI agents flooding races, school board, and mayoral races with fake news outlets coordinated at a level current identification systems can’t handle.
“The last war, which was in 2024, we won,” he said. “The next battle is coming.”
In leadership: Etzioni said AI adoption is not something leaders can delegate to the CIO or general counsel. His three ideas:
- Use AI yourself, whether you’re a CEO or a caretaker.
- Create incentive structures that encourage your team to check it out.
- Don’t use AI to do existing work immediately; see what is possible only with AI.
“The real gold,” he said, “is when you get AI to do new things that we haven’t done before.”
Listen to the full interview on the GeekWire Podcast above, or subscribe to GeekWire on Apple Podcasts, Spotify, or wherever you listen.
