He’s previously said this information will be used to train X’s artificial intelligence chatbot Grok on how to interpret them efficiently.

Earlier this month, Elon Musk reposted a video on X of himself talking about uploading medical data to Grok, saying: “Try it!”

“You can upload your X-rays or MRI images to Grok and it will give you a medical diagnosis,” Musk said in the video, which was uploaded in June. “I have seen cases where it’s actually better than what doctors tell you.

In 2024, Musk said medical images uploaded to Grok would be used to train the bot.

“This is still early stage, but it is already quite accurate and will become extremely good,” Musk wrote on X. “Let us know where Grok gets it right or needs work.”

Musk also claimed in his response Grok saved a man in Norway by diagnosing a problem his doctors failed to notice. The X owner was willing to upload his own medical information to his bot. 

“I did an MRI recently and submitted it to Grok,” Musk said in an episode of the Moonshots with Peter Diamandis podcast released on Tuesday. “None of the doctors nor Grok found anything.”

Musk did not disclose in the podcast why he received an MRI. XAI, which owns X, told Fortune in a statement: “Legacy Media Lies.”

Grok is facing some competition in the AI health space. This week OpenAI launched ChatGPT Health, an experience within the bot feature that allows users to securely connect medical records and wellness apps like MyFitnessPal and Apple Health. The company said it would not train the models using personal medical information.

AI chatbots have become a ubiquitous source of medical information for people. OpenAI reported this week 40 million people seek health information from the model, 55% of which used to bot to look up or better understand symptoms.

Dr. Grok will see you now

So far, Grok’s ability to detect medical abnormalities have been mixed. The AI successfully analyzed blood test results and identified breast cancer, some users claimed. But it also grossly misinterpreted other pieces of information, according to physicians who responded to some of Musk’s about Grok’s ability to interpret medical information. In one instance, Grok mistook a “textbook case” of tuberculosis for a herniated disk or spinal stenosis. In another, the bot mistook a mammogram of a benign breast cyst for an image of testicles.

A May 2025 study found that while all AI models have limitations in processing and predicting medical outcomes, Grok was the most effectively compared to Google’s Gemini and ChatGPT-4o when determining the presence of pathologies in 35,711 slices of brain MRI.

“We know they have the technical capability,” Dr. Laura Heacock, associate professor at the New York University Langone Health Department of Radiology, wrote on X. “Whether or not they want to put in the time, data and [graphics processing units] to include medical imaging is up to them. For now, non-generative AI methods continue to outperform in medical imaging.”

The problems with Dr. Grok

Musk’s lofty goal of training his AI to make medical diagnoses is also a risky one, experts said. While AI has increasingly been used as a means to make complicated science more accessible and create assistive technologies, teaching Grok to use data from a social media platform presents concerns about both Grok’s accuracy and user privacy.

Ryan Tarzy, CEO of health technology firm Avandra Imaging, said in an interview with Fast Company asking users to directly input data, rather than source it from secure databases with de-identified patient data, is Musk’s way of trying to accelerate Grok’s development. Also, the information comes from a limited sample of whoever is willing to upload their images and tests—meaning the AI is not gathering data from sources representative of the broader and more diverse medical landscape.

Medical information shared on social media isn’t bound by the Health Insurance Portability and Accountability Act (HIPAA), the federal law that protects patients’ private information from being shared without their consent. That means there’s less control over where the information goes after a user chooses to share it.

“This approach has myriad risks, including the accidental sharing of patient identities,” Tarzy said. “Personal health information is ‘burned in’ too many images, such as CT scans, and would inevitably be released in this plan.”

The privacy dangers Grok may present aren’t fully known because X may have privacy protections not known by the public, according to Matthew McCoy, assistant professor of medical ethics and health policy at the University of Pennsylvania. He said users share medical information at their own risk.

“As an individual user, would I feel comfortable contributing health data?” he previously told the New York Times. “Absolutely not.”

A version of this story originally published on Fortune.com on Nov. 20, 2024.

More on AI and health:

The mounting unrest comes as Tehran’s piecemeal efforts to address an economic crisis have done little to appease Iranians. Since protests began late last month, the government has offered words of sympathy, fired the central bank’s chief, and announced plans to provide most people with a monthly payment of about 1 million Iranian tomans—equivalent to $7.

Instead of spending $10 billion annually to subsidize imports, that money will instead go directly to 80 million Iranians in the form of credit to buy certain goods.

But the $7 monthly payments offer little relief to beleaguered consumers who are suffering from food inflation of 64%. It’s made worse by a 60% crash in the currency’s value since June, when Iran and Israel fought a 12-day war that was capped by the U.S. bombing of Tehran’s nuclear facilities.

Now, what began as a protest among merchants in Tehran’s bazaars has spread to students as well as Iran’s working and middle classes all across the country.

The security forces that keep the regime in power have not escaped hardship either. While human rights groups estimate hundreds have died from the government’s crackdown, Iranians say it’s not as severe as it could be.

“Security and law enforcement people are facing the same economic issues and high prices, themselves,” a protester in Tehran told the New York Times. “They are not fighting back wholeheartedly.”

Meanwhile, Trump has threatened Iran if the regime kills protesters and doubled down on that Friday, when he said the U.S. would “start shooting” if authorities fired on demonstrators.

With the violence worsening, Trump is looking at ways to follow through. Reports said that administration officials have already started discussing options to attack Iran again. On Sunday, sources told the Wall Street Journal that Trump will be briefed on Tuesday with Secretary of State Marco Rubio, Defense Secretary Pete Hegseth and Joint Chiefs Chair Gen. Dan Caine also due to attend.

In addition to military strikes, other options include boosting antigovernment sources online, cyber attacks, and more economic sanctions, the report said.

But the Journal added that the Pentagon hasn’t sent any forces to the region and that the redeployment of the a USS Gerald R. Ford aircraft carrier to South America means there are none in the Middle East or Europe now.

The U.S. raid on Venezuela last week to capture Nicolas Maduro could weigh on military considerations for Iran as a large armada of Navy ships remain in the Caribbean and continue to enforce a “quarantine” on the country’s oil.

But Trump has shown his appetite for more foreign intervention hasn’t abated, even as the reality of a years-long commitment to rebuild Venezuela’s shattered oil industry sets it.

On Sunday, he sent another warning via social media to Cuba, which had benefited from economic assistance when Maduro was in power but is now feeling more strains.

“THERE WILL BE NO MORE OIL OR MONEY GOING TO CUBA – ZERO!” Trump said in a post. “I strongly suggest they make a deal, BEFORE IT IS TOO LATE.”

Over the course of 2023 and into the first quarter of 2024, Vaughan told Fortune, IgniteTech replaced hundreds of employees, declining to disclose a specific number. “That was not our goal,” he told Fortune. “It was extremely difficult … But changing minds was harder than adding skills.” It was, by any measure, a brutal reckoning—but Vaughan insists it was necessary, and said he’d do it again.

For Vaughan, the writing on the wall was clear and dramatic.

“In early 2023, we saw the light,” he told Fortune in an August 2025 interview, adding he believed every tech company was facing a crucial inflection point around adoption of artificial intelligence. “Now I’ve certainly morphed to believe that this is every company, and I mean that literally every company, is facing an existential threat by this transformation.”

Where others saw promise, Vaughan saw urgency—believing failing to get ahead on AI could doom even the most robust business. He called an all-hands meeting with his global remote team. Gone were the comfortable routines and quarterly goals. Instead, his message was direct: Everything would now revolve around AI. “We’re going to give a gift to each of you. And that gift is tremendous investment of time, tools, education, projects … to give you a new skill,” he explained. The company began reimbursing for AI tools and prompt-engineering classes, and even brought in outside experts to evangelize.

“Every single Monday was called ‘AI Monday,’” Vaughan said, with his mandate for staff that they could work only on AI. “You couldn’t have customer calls; you couldn’t work on budgets; you had to only work on AI projects.” He said this happened across the board, not just for tech workers, but also for sales, marketing, and everybody else at IgniteTech. “That culture needed to be built. That was the key.”

This was a major investment, he added: 20% of payroll was dedicated to a mass-learning initiative, and it failed because of mass resistance, even sabotage. Belief, Vaughan discovered, is a hard thing to manufacture.

“In those early days, we did get resistance, we got flat-out, ‘Yeah, I’m not going to do this’ resistance,” he said. “And so we said goodbye to those people.”

The pushback: white collar resistance

Vaughan was surprised to find it was often the technical staff, not marketing or sales, who dug in their heels. They were the “most resistant,” he said, voicing various concerns about what the AI couldn’t do, rather than focusing on what it could. The marketing and salespeople were enthused by the possibilities of working with these new tools, he added.

This friction is borne out by broader research. According to the 2025 enterprise AI adoption report by Writer, an agentic AI platform for enterprises, one in three workers say they’ve “actively sabotaged” their company’s AI rollout—a number that jumps to 41% of millennial and Gen Z employees. This can take the form of refusing to use AI tools, intentionally generating low-quality outputs, or avoiding training altogether. Many act out because of fears that AI will replace their jobs, while others are frustrated by lackluster AI tools or unclear strategy from leadership.

Writer’s chief strategy officer Kevin Chung told Fortune the “big eye-opening thing” from this survey was the human element of AI resistance.

“This sabotage isn’t because they’re afraid of the technology,” he said. “It’s more like there’s so much pressure to get it right, and then when you’re handed something that doesn’t work, you get frustrated.”

He added Writer’s research shows workers often don’t trust where their organizations are headed.

“When you’re handed something that isn’t quite what you want, it’s very frustrating, so the sabotage kicks in, because then people are like, ‘Okay, I’m going to run my own thing. I’m going to go figure it out myself.’” You definitely don’t want this kind of “shadow IT” in an organization, he added.

Vaughan said he didn’t want to force anyone.

“You can’t compel people to change, especially if they don’t believe,” he said, adding belief was really the thing he needed to recruit for.

Company leadership ultimately realized they’d have to launch a massive recruiting effort for what became known as “AI innovation specialists.” This applied across the board: to sales, finance, marketing, and elsewhere. Vaughan said this time was “really difficult” as things inside the company were “upside down … We didn’t really quite know where we were or who we were yet.”

A couple of key hires helped, starting with the person who became IgniteTech’s chief AI officer, Thibault Bridel-Bertomeu. That led to a full reorganization of the company that Vaughan called “somewhat unusual.” Essentially, every division came to report into the AI organization, regardless of domain.

This centralization, Vaughan said, prevented duplication of efforts and maximized knowledge sharing—a common struggle in AI adoption, where Writer’s survey shows 71% of the C-suite at other companies say AI applications are being created in silos and nearly half report their employees have been left to “figure generative AI out on their own.”

No pain, no gain?

In exchange for this difficult transformation, IgniteTech reaped extraordinary results. By the end of 2024, the company had launched two patent-pending AI solutions, including a platform for AI-based email automation (Eloquens AI), with a radically rebuilt team.

Financially, IgniteTech remained strong. Vaughan disclosed the company, which he said was in the nine-figure revenue range, finished 2024 at “near 75% Ebitda”—all while completing a major acquisition, Khoros.

“You multiply people … give people the ability to multiply themselves and do things at a pace,” he said, touting the company’s ability to build new customer-ready products in as little as four days, an unthinkable timeline in the old regime. In the months since, Vaughan told Fortune in an early 2026 statement, the company has only kept growing its headcount, recruiting globally for AI Innovation Specialists across every function, from marketing to sales to finance to engineering to support.

What does Vaughan’s story say for others? On one level, it’s a case study in the pain and payoff of radical change management. But his ruthless approach arguably addresses many challenges identified in the Writer survey: lack of strategy and investment, misalignment between IT and business, and the failure to engage champions who can unlock AI’s benefits.

The ‘boy who cried wolf’ problem

To be sure, IgniteTech is far from alone in wrestling with these challenges. Joshua Wöhle is the CEO of Mindstone, a firm that provides AI upskilling services to workforces, training hundreds of employees monthly at companies including Lufthansa, Hyatt, and NBA teams. He recently discussed the two approaches described by Vaughan—upskilling and mass replacement—in an appearance on BBC Business Today.

Wöhle contrasted the recent examples of Ikea and Klarna, arguing the former’s example shows why it’s better to “reskill” existing employees. Klarna, a Swedish buy-now, pay-later firm, drew considerable publicity for a decision to reduce members of its customer support staff in a pivot to AI, only to rehire for the same roles.

“We’re near the point where [AI is] more intelligent than most people doing knowledge work. But that’s precisely why augmentation beats automation,” Wöhle wrote on LinkedIn.

A representative for Klarna told Fortune the company did not lay off employees, but has instead adopted several approaches to its customer service, which is managed by outsourced customer service providers who are paid according to the volume of work required. The launch of an AI customer service assistant reduced the workload by the equivalent of 700 full-time agents—from roughly 3,000 to 2,300—and the third-party providers redeployed those 700 workers to other clients, according to Klarna. Now that the AI customer service agent is “handling more complex queries than when we launched,” Klarna says, that number has fallen to 2,200. Klarna says its contractor has rehired just two people in a pilot program designed to combine highly trained human support staff with AI to deliver outstanding customer service. 

In an interview with Fortune, Wöhle said one client of his has been very blunt with his workers, ordering them to dedicate all Fridays to AI retraining, and if they didn’t report back on any of their work, they were invited to leave the company.

He said it can be “kinder” to dismiss workers who are resistant to AI: “The pace of change is so fast that it’s the kinder thing to force people through it.” He added he used to think if he got all workers to really love learning, then that could help Mindstone make a real difference, but he discovered after training literally thousands of people that “most people hate learning. They’d avoid it if they can.”

Wöhle attributed much of the AI resistance in the workforce to a “boy who cried wolf” problem from the tech sector, citing NFTs and blockchain as technologies that were billed as revolutionary but “didn’t have the real effect” that tech leaders promised.

“You can’t really blame them” for resisting, he said. Most people “get stuck because they think from their work flow first,” he added, and they conclude AI is overhyped because they want AI to fit into their old way of working. “It takes a lot more thinking and a lot more kind of prodding for you to change the way that you work,” but once you do, you see dramatic increases. A human can’t possibly keep five call transcripts in their head while you’re trying to write a proposal to a client, he offers, but AI can.

Ikea echoed Wöhle when reached for comment, saying its “people-first AI approach focuses on augmentation, not automation.” A spokesperson said Ikea is using AI to automate tasks, not jobs, freeing up time for value-added, human-centric work.

The Writer report notes companies with formal AI strategies are far more likely to succeed, and those who heavily invest in AI outperform their peers by a large margin. But as Vaughan’s experience shows, investment without belief and buy-in can be wasted energy. “The culture needed to be built. Ultimately, we ended up having to go out and recruit and hire people that were already of the same mind. Changing minds was harder than adding skills.”

From the vantage point of early 2026, Vaughan reflected in a statement to Fortune, monthly all-hands meetings look nothing like they used to: “We killed the format of reviewing goals and metrics. Now teams demo what they built.” He wanted to stress something else: Despite the drastic actions he took to restructure, he still doesn’t think he’s ahead of the curve.

“We’re just not getting run over from behind yet,” he said. “The pace of change in AI is relentless. If we don’t keep pushing, keep learning every single day, we’re toast.”

For Vaughan, there’s no ambiguity. Would he do it again? He doesn’t hesitate: He’d rather endure months of pain and build a new, AI-driven foundation from scratch than let an organization drift into irrelevance.

“This is not a tech change. It is a cultural change, and it is a business change,” he said, adding he doesn’t recommend others follow his lead and swap out 80% of their staff.

“I do not recommend that at all,” he said. “That was not our goal. It was extremely difficult.”

But at the end of the day, he added, everybody’s got to be in the same boat, rowing in the same direction. Otherwise, “we don’t get where we’re going.”

A version of this story was published on Fortune.com on August 17, 2025.

More on AI in the workplace: