When OpenAI declared a “code red” this month to refocus its teams on competing with Google, I couldn’t help but think back to December three years ago when the companies’ roles were reversed. Google was the one blasting the sirens to catch up to OpenAI. What followed the next month, in January 2023, were the first sweeping layoffs in Google’s history. “A difficult decision to set us up for the future,” as the company described it at the time.
I wonder whether the ChatGPT developer could make similar workforce cuts early next year. This speculation inspired me to come up with a whole set of predictions about what might come in the year ahead. Here’s a look at six of the ideas, fine-tuned with the real intelligence of WIRED colleagues.
Data Center Disinformation
Communities across the world are fighting the construction of data centers. In the US, many activists are organizing on social media using tools such as Facebook Groups. The Chinese and Russian governments continue to exploit social media to disseminate disinformation masquerading as real news and authentic opinion. Slowing data center development in the US would be a boon for China and Russia, which are both seeking to surpass the US in industrial and military AI capabilities.
Austin Wang, a researcher at the nonprofit think tank RAND who has studied China-controlled propaganda farms, says there’s no signs of concerning activity right now. “Many newly established anti-data-center pages seem controlled by real US citizens so far,” Wang says.
But as the anti-data-center fervor picks up, China and Russia could try to pile on to the grassroots organizing. And the work has gotten even easier thanks to AI that can quickly generate images and videos to rile up people on social media.
Robot Demos Everywhere
In 2026, tech conferences from the Consumer Electronics Show to Amazon’s hardware event will likely be buzzing about AI-powered robots. Google and other big tech companies have spent years trying to train robots to handle household tasks through repeated practice. But now there’s a fresh round of hype. The type of AI models used in services such as ChatGPT and Gemini are being integrated into robots in hopes that they will handle chores, like folding clothes, with less training and greater accuracy.
This past September, Google released a video of a robot sorting trash, compost, and recycling in response to a user’s voice commands. When Google executives take to the stage at the company’s next I/O conference, I expect them to prompt a robot to take on tasks such as, for example, sliding a pizza into a type of oven it’s never encountered before and, while it cooks, retrieving a half-full Diet Coke from the back of a crowded fridge.
Barak Turovsky, the recently departed chief AI officer at General Motors and a former leader in Google’s AI division, says advancements in robots’ capabilities are possible because large language models can understand a dishwasher manual, learn how to operate a dishwasher from watching a video, and comprehend how to grab a specific part by deciphering a drawing. “The next frontier for large language models is the physical world,” he says.
To be clear, the showcases of next year will be demonstrations. Selling technology that could physically wreck people’s homes if they err will first require additional testing.
The Bubble Deflates
2025 began with China’s DeepSeek showing the world that you don’t need a ton of cutting-edge GPUs to make a solid AI system. This prompted a fleeting stock market selloff over fears that chip sales would tank. The fears didn’t materialize. But next year may bring a bigger, once-again ephemeral dip. Leading AI companies may need a reset to doubledown on successful investments and trim struggling ventures after a period of torrid growth, and their moves could end up being cast by tech pundits as a sign of overspending on AI data centers and researchers.
OpenAI has pentupled during the past two years to about 4,500 employees, according to company data. It is fighting many battles—not just against Google—and expanding into many new facets, like designing its own chips alongside Broadcom, so the personnel growth could be warranted. But does it still have the best people in the best roles? Newly onboarded management may see things differently, and that’s why the 10-year-old organization’s first major layoffs may be coming next year. If that happens, other AI labs could follow OpenAI’s lead with their own restructuring.
OpenAI spokesperson Jason Deutrom says, “ChatGPT may be everywhere, but we’re still a relatively small team” and that it is “excited to keep hiring and building more stuff people love in 2026.”
Some tech companies will try to launch initial public offerings to cash in on peak valuations before the AI-fueled stock market sours. Analysts who study IPO prospects expect a brimming pipeline to gusher in 2026, with chat service Discord, payments processor Stripe, and cloud platform Databricks among perennial rumored names. That said, preparing for an IPO is a challenging task, and nailing the timing to capitalize on a bullish mood on Wall Street is even more difficult. Companies that miss the coveted “window” could add to the wave of workforce cuts.
Training Work Agents
For years, companies have installed “bossware” on workers’ computers to monitor for potentially inappropriate behavior. I am not looking forward to a possible reality of the coming year: Surveillance software aimed at recording employees’ work to train AI agents to automate some of the tasks.
Agentic AI that can reduce manual labor by responding to, say, customer service queries is already spreading. These tools are often trained using data that is synthetically generated by computers or gathered from monitoring the activity of click workers who are paid to simulate work.
But as businesses look to automate complex jobs and tasks that are simple but take a lot of steps, they will need training data more specific to their working environments. Cue the software to slurp up user click, scroll, and typing activity. “The capabilities are there and emerging where one can see this happening,” says Wilneida Negrón, a workers’ rights activist who has studied employment tech.
The situation could add to worker concerns about job loss. And there’s also the fear that one of these tools inadvertently captures more than it should, including personal information, and accidentally makes it accessible to colleagues.
Always On, Always Danger
AI gadgets like necklaces with always-on microphones that are capable of tracking a user’s every word turned out to be duds in 2025. But AI software that listens to video calls and other audio interactions on the computer were a surprise hit.
One option, Granola, uses AI to generate meeting notes without storing a permanent audio recording. “Their output is a relevant, well-organized and truly useful outline of what took place in a long complex call with lots of participants,” says Javier Soltero, the former head of Google Workspace software. The rub is that Granola can operate without other call participants knowing, though the company advises seeking consent. The company’s “argument seems to be ‘Well, you could be taking notes and not feel compelled to tell people about it,’” Soltero says.
The proliferation of these services is prompting new questions about digital etiquette, accessibility, and the law. I am betting these issues will come to the forefront through at least one major data breach or privacy lawsuit in 2026. “The question of how AI systems affect third parties, other than the user who’s actually engaging with the system, is important—and agentic AI is likely to make this even more pressing,” says Alicia Solow-Niederman, associate professor of law at George Washington University.
Some in Silicon Valley are adapting. “I’ve taken the view personally that there’s a lot of instances in which I’m not aware things are being recorded, and that’s scary,” says Talia Goldberg, an investor at venture capital firm Bessemer. Steve Jang, founder and managing partner of Kindred Ventures, says, “All of AI sits in this gray uncertain area in terms of protocol and usage.”
Always-recording wearables and apps seem here to stay—and it’s not all bad. They’re a helpful tool for people who are not only in endless meetings but deaf and reliant on assistive technology. Companies just may need to put in some more guardrails next year.
Robotaxi Takeover, Without Incident
US robotaxi services are poised for a big expansion in 2026. Chiefly, Waymo expects to provide more than 1 million rides per week by the end of next year, up from hundreds of thousands. The Google sibling company, which is reportedly fundraising for $15 billion, will potentially increase service to up to about 25 cities from five, including going abroad to London and Tokyo. Tesla and Amazon-owned Zoox also have announced plans to increase their offerings.
A popular prediction would be that this buildup in driverless ride-hailing services in the US will lead to the industry’s first deadly accident for which the computer is at fault. Self-driving cars are in dozens of accidents per month, according to estimates from federal authorities. But federal, state, and industry data show robotaxis are rarely the cause, and only a small percentage of the incidents have resulted in deaths to living beings.
The number of robotaxis on the road remains limited. While some now serve highways, speeds are generally slow. More likely, accidents caused by human drivers will pile up. As will accidents caused by pseudo-autopilot systems that people overly rely on to maneuver their cars. Robotaxis, though, will be incentivized to play it safe and stay horror-free in 2026.
The post 6 Scary Predictions for AI in 2026 appeared first on Wired.
