Most corporate rivalries start over money. This one started over a disagreement about how careful the world's most powerful technology should be, and it has been getting messier in public ever since.
By August 2026, two companies built by the same small group of people were running attack ads against each other during the Super Bowl, quietly funding opposing candidates in a New York congressional race, and standing next to each other at a world summit without touching hands.
Here are the 15 moments that turned a research disagreement into what the industry is now openly calling an AI civil war.
Table of Contents
- The VP of Research Who Walked Out
- The Founding Family That Split in Two
- Two Chatbots, One Bloodline
- The Billboard Shade Heard Around San Francisco
- The Super Bowl Ad War: "Deception," "Betrayal," "Treachery," "Violation"
- Altman's "Orwellian Doublespeak" Clapback
- The Handshake That Wasn't
- The Pentagon Snub That Became a Ban
- OpenAI's 24-Hour Pentagon Turnaround
- The $27 Million Congressional Proxy War
- The Winner Who Told Both Sides to Get Lost
- Microsoft's Exclusivity Ends, Amazon Walks In
- The Enterprise Land Grab
- Dueling IPO Filings, One Week Apart
- The Privacy One-Upmanship of August 2026
Note: Every fact below is sourced to real, dated reporting. Nothing here is speculation dressed up as news.
1. The VP of Research Who Walked Out
Before he ran a rival company, Dario Amodei ran research at OpenAI, rising to vice president of research, where he focused heavily on safety.
- ✅ He played an instrumental role in launching GPT-2 and GPT-3.
- ✅ He co-invented reinforcement learning from human feedback (RLHF), the training method that now underpins how nearly every major chatbot behaves.
- ✅ Colleagues who worked alongside him during this period would later follow him out the door entirely.
His departure wasn't a quiet resignation. It was the first crack in what would become a years-long fracture, and it's a pattern the industry has kept repeating since, right down to Meta's record-breaking talent raid on rival labs.
More from the AI talent wars: how one researcher's departure turned into a $200 million bidding war between the biggest labs in the industry.
2. The Founding Family That Split in Two
In 2021, Amodei left for good, and he didn't leave alone. He co-founded Anthropic with his sister and fellow OpenAI alum Daniela Amodei, along with a dozen other former OpenAI researchers and staff. The group's stated reason for leaving was concern that OpenAI was prioritizing speed and scale over safety as it raced to commercialize its models.
That detail matters more than a typical "startup poaches talent" story. This wasn't outsiders competing with an incumbent. This was people who built the technology together, splitting over how it should be built going forward, and then becoming each other's fiercest competitors, the same safety-first instinct that later fed the broader chorus of researchers warning about where the technology is headed.
Related reading: 12 warnings from the scientists who helped build the AI systems now worrying them.
3. Two Chatbots, One Bloodline
The products that came out of that split now define the industry.
- ✅ OpenAI released ChatGPT in late 2022, turning the company into a household name almost overnight.
- ✅ Anthropic followed with its first version of Claude in 2023, positioning it from day one as the "safety-first" alternative.
Every comparison between the two chatbots since then carries the weight of that origin story. It's not just Coke versus Pepsi. It's two former colleagues each betting their careers on a different answer to the same question, a fight that has only escalated as both companies now battle Google for dominance.
If you want the wider picture: how the Google-OpenAI rivalry is reshaping the entire AI market.
4. The Billboard Shade Heard Around San Francisco
Long before the feud went fully public, Anthropic was already taking quiet shots. In May 2025, billboards went up around San Francisco reading "AI that you can trust" and "The one without all the drama." Anthropic never named OpenAI directly, but the timing and phrasing were hard to miss. Many in the industry read it as a direct reference to the chaotic November 2023 boardroom coup that briefly ousted Sam Altman from OpenAI, followed by a wave of high-profile staff departures.
It was the corporate equivalent of a subtweet, and it set the tone for everything that followed: neither company would say the other's name outright, but everyone would know exactly who they meant.
5. The Super Bowl Ad War: "Deception," "Betrayal," "Treachery," "Violation"
By February 2026, Anthropic stopped being subtle. The company spent millions on a four-ad campaign called "A Time and a Place," which aired during the Super Bowl. Each spot opened with a single loaded word before showing ordinary people getting helpful AI assistance suddenly interrupted by an unrelated ad.
- ✅ Ad one: "Deception"
- ✅ Ad two: "Betrayal"
- ✅ Ad three: "Treachery"
- ✅ Ad four: "Violation"
The tagline tied it all together: "Ads are coming to AI. But not to Claude."
None of the ads mentioned OpenAI by name. They didn't need to. OpenAI had already signaled plans to bring advertising into ChatGPT, and the timing made the target obvious to anyone paying attention. OpenAI didn't sit out the ad slot either. Its own Super Bowl spot leaned earnest instead of combative, spotlighting its coding tool Codex and framing the company around the idea that anyone can build with AI.
6. Altman's "Orwellian Doublespeak" Clapback
Sam Altman didn't let the Super Bowl ads slide. He pushed back publicly, calling Anthropic's campaign clearly dishonest and describing it as Orwellian doublespeak. He argued that OpenAI's ad-supported model exists to bring AI to billions of people who can't afford a subscription, not to trick anyone.
- ❌ Altman rejected the idea that OpenAI would run ads the way Anthropic's commercials depicted them.
- ❌ He said OpenAI knew its users would reject that kind of experience outright.
Anthropic's president, Daniela Amodei, went on Good Morning America afterward and insisted the ads weren't really about OpenAI, or any company other than Anthropic itself. Almost nobody in the industry took that at face value.
7. The Handshake That Wasn't
The rivalry moved from advertising to body language just weeks later. On February 19, 2026, Indian Prime Minister Narendra Modi hosted the AI Impact Summit in New Delhi, gathering the world's top AI executives on stage. When Modi asked the group to join hands and raise them in a show of unity, everyone complied, including Google's Sundar Pichai and Meta's AI chief.
Everyone except Sam Altman and Dario Amodei.
- ✅ Standing right next to each other, both men raised separate fists instead of clasping hands.
- ✅ The moment was captured on video and went viral almost immediately.
- ✅ Altman later told CNBC he was simply "confused" by what Modi wanted them to do, calling it an unintentional mix-up rather than a snub.
Whether it was accidental or not, the internet had already made up its mind. The image became shorthand for the entire feud: two men who used to work together, now unable to even fake a moment of unity on camera.
8. The Pentagon Snub That Became a Ban
While the public feud played out in ads and photo ops, a quieter and more consequential fight was happening over defense contracts. Anthropic reportedly declined a Pentagon deal, citing concerns tied to autonomous weapons systems and surveillance use cases that conflicted with its safety-first positioning. On March 3, 2026, Anthropic received a formal "supply chain risk" designation from the Department of War, a classification that effectively barred Claude from use across US federal systems.
This wasn't a symbolic dispute. It was a company's safety principles colliding directly with a government contract, and the government responding by shutting the door.
Update: the story didn't end there. A federal judge later ruled the designation unlawful, finding it amounted to unconstitutional retaliation against Anthropic and a violation of its due process rights. The Pentagon is expected to appeal, so the fight over Claude's federal access isn't fully resolved even now.
Same standoff, bigger picture: inside the broader AI arms race and why governments are picking sides.
9. OpenAI's 24-Hour Pentagon Turnaround
OpenAI moved almost immediately. Less than a day after Anthropic's ban became public, Sam Altman announced that OpenAI had signed its own deal to deploy AI models on classified US Department of War networks. The agreement included specific guardrails: a ban on domestic mass surveillance and a requirement for human control over any lethal force decisions.
- ✅ OpenAI framed the deal as responsible engagement with national security rather than a rejection of safety principles.
- ❌ Critics online saw it differently, with a wave of users publicly announcing plans to cancel their ChatGPT subscriptions in protest.
The contrast was stark. One company walked away from defense work over safety concerns and got frozen out of federal systems entirely. The other stepped directly into the gap, guardrails attached, and picked up the business its rival had turned down.
For more on how military AI contracts are reshaping the industry: the warning signs behind an AI-fueled arms race.
It set up the next phase of the rivalry, where the fight stopped being about ads and photo ops and started being about actual dollars, actual contracts, and actual political influence.
10. The $27 Million Congressional Proxy War
The feud didn't stay confined to tech. It spilled directly into American politics. In a Manhattan congressional primary, two AI-aligned political forces poured a combined $27 million into deciding who would represent New York's 12th district. State assemblyman Alex Bores became the unlikely center of the fight, backed heavily by Anthropic-aligned money.
The roots of the spending war trace back further:
- In August 2025, a super PAC called Leading the Future launched with more than $100 million in initial funding, backed by investors including Andreessen Horowitz, OpenAI co-founder Greg Brockman, and Perplexity.
- Industry analysts framed the PAC's creation as a direct response to Anthropic's growing involvement in pushing AI-related legislation.
- Anthropic's co-founders had split from OpenAI back in early 2021 specifically over safety disagreements, and that same philosophical divide was now playing out through campaign finance filings instead of research papers.
The message from the pro-industry side was blunt, according to political strategists tracking the race: push for AI regulation without the industry's input, and you become a target.
11. The Winner Who Told Both Sides to Get Lost
Here's the twist nobody in either camp saw coming. The $27 million didn't buy the outcome either side wanted. Democrat Micah Lasher won the primary, and no AI company had spent significant money trying to elect him. His victory speech didn't thank Silicon Valley. It called it out directly.
- ❌ "I have some news for the two big AI companies who've taken such an unusual interest in who won this congressional seat," Lasher told the crowd on primary night.
- ❌ He said he wouldn't be taking cues from either company when it came to protecting kids, jobs, or the environment.
For an industry used to buying influence quietly, watching a candidate win specifically by rejecting both sides in public was a wake-up call. The AI civil war had officially become expensive enough, and public enough, to become a liability for everyone involved.
12. Microsoft's Exclusivity Ends, Amazon Walks In
While Anthropic and OpenAI were busy fighting each other, one of the biggest shifts in the entire relationship came from a third party quietly stepping back. By April 27, 2026, OpenAI had formally ended its seven-year exclusive cloud partnership with Microsoft. The original 2019 deal, built on a $1 billion investment, had three pillars: Microsoft as OpenAI's exclusive cloud provider, joint supercomputing development, and a license to commercialize OpenAI's pre-AGI technology. It had no fixed end date, running instead "until AGI is achieved," a term that was never precisely defined.
That arrangement started cracking after OpenAI released its o1 reasoning model in September 2024, which triggered disputes over intellectual property and exposed chronic inference-performance problems on Azure that reportedly pushed even well-funded startups elsewhere.
Within a day of the exclusivity ending, Amazon's models went live on AWS Bedrock as part of the up-to-$50 billion deal Amazon had committed to OpenAI, using Amazon's custom Trainium chips. The renegotiated Microsoft deal stripped away exclusivity, revenue share, and the vague "AGI escape clause" that had defined the original partnership.
- ✅ OpenAI gained the freedom to sell on any cloud provider it wanted.
- ✅ Microsoft kept Copilot deeply embedded in Microsoft 365, Windows, and GitHub, losing exclusivity but not relevance.
It reshaped the entire AI infrastructure map overnight, and it happened almost entirely separate from the public Anthropic feud, a reminder that OpenAI was fighting battles on more than one front at once.
13. The Enterprise Land Grab
With the cloud relationship settled, OpenAI turned its full attention back to Anthropic, this time over enterprise customers. By April 2026, OpenAI was actively mobilizing consulting partners to help large companies deploy and scale Codex, its coding tool, specifically to claw back enterprise market share it had lost to Anthropic's Claude.
- OpenAI leaned on its compute advantage and consulting partnerships to grow user numbers aggressively.
- Anthropic focused on protecting its margins rather than chasing volume, betting that enterprise customers valued stability over speed.
Industry analysts framed the split strategy as a preview of what would matter most heading into 2026's biggest business event for both companies: going public. Whoever won the enterprise fight would walk into that moment with more leverage, and whoever lost would have to explain to investors exactly why.
14. Dueling IPO Filings, One Week Apart
After years of ads, snubs, and political spending, the rivalry moved to Wall Street. On June 1, 2026, Anthropic announced it had confidentially filed a draft S-1 with the SEC, the first official step toward going public. One week later, on June 8, OpenAI did the exact same thing.
- ✅ Neither company has confirmed whether they'll actually list in the same window.
- ✅ Both were playing catch-up to Elon Musk's xAI, which made its Nasdaq debut in June 2026, a moment that fits into the wider superintelligence race Musk himself has been vocal about.
Days after the S-1 announcements, Dario Amodei sat down for a lengthy interview with Bloomberg's "The Circuit with Emily Chang." Asked directly why he left OpenAI, he didn't soften the answer. "At the end of the day, why argue with someone when you don't have the same vision and you don't trust them?" Amodei said, a line widely read as an unmistakable jab at his former boss.
He also made clear the rivalry wouldn't be settled in interviews or ad campaigns. Amodei said the public itself would ultimately decide which company comes out ahead, through the products they choose to use and the money they choose to invest.
More on the money side of this race: what Elon Musk has said publicly about the superintelligence race his own companies are now part of.
15. The Privacy One-Upmanship of August 2026
As recently as this month, the two companies are still finding new fronts to compete on. OpenAI announced a preview of a new offering called Private Safety Processing, an automated system designed to monitor for potential misuse of its models while retaining none of the customer's underlying data. The framing was direct: a privacy-first approach specifically built to one-up Anthropic's existing data-retention policy.
The move highlights a real tension across the entire AI industry, not just between these two companies: how do you watch for abuse of powerful models without compromising the privacy you promise enterprise customers? Anthropic hasn't publicly responded to OpenAI's new offering yet, but given the pattern of the last eighteen months, a countermove seems less like a possibility and more like a formality.
Where the Rivalry Stands Now?
Two companies. One shared origin story. Fifteen very different ways of showing the world they no longer agree on much of anything.
What makes this rivalry different from a typical tech feud is the history underneath it. This isn't Coke sniping at Pepsi, or Uber sniping at Lyft. This is a group of people who built some of the most consequential technology of this century together, split apart over how carefully it should be handled, and then spent the better part of five years proving to the public exactly how deep that disagreement runs.
Neither company shows signs of backing down. If anything, as both march toward IPOs that will put their rivalry under even more public scrutiny, 2026 has made one thing clear: this fight is just getting started.
If you're following where this all leads next: our predictions for what 2027 looks like across the AI industry.
Frequently Asked Questions (FAQ)
Did OpenAI actually split into two separate companies?
No. OpenAI has not split into two companies. The "civil war" refers to the rivalry between OpenAI and Anthropic, a separate company founded in 2021 by former OpenAI researchers, including Dario and Daniela Amodei, after disagreements over AI safety.
Why did Dario Amodei leave OpenAI?
Amodei worked at OpenAI from 2016 to 2020 as vice president of research. He left, along with a dozen other researchers, because of concerns that OpenAI was prioritizing speed and scale over safety as it commercialized its models.
What was the Super Bowl ad war about?
In February 2026, Anthropic ran a four-ad Super Bowl campaign criticizing plans to bring advertising into AI chatbots, widely seen as an unnamed jab at OpenAI. Sam Altman publicly called the ads "dishonest" and "Orwellian doublespeak" in response.
Why did Anthropic get banned from federal systems, and is it still banned?
Anthropic reportedly declined a Pentagon deal over concerns tied to autonomous weapons and surveillance. In response, it received a formal "supply chain risk" designation in March 2026 that barred Claude from US federal systems. A federal judge later ruled that designation unlawful, though the Pentagon is expected to appeal, so the matter isn't fully settled.
Are Anthropic and OpenAI both going public?
Both companies confidentially filed draft S-1 paperwork with the SEC in June 2026, Anthropic on June 1 and OpenAI on June 8, signaling plans to eventually go public, though neither has confirmed a final listing date.
Conclusion
Two companies built by the same people, now betting billions on opposite answers to the same question. That's not a rivalry. That's a family argument the entire world happens to be watching, and the next round hasn't even started yet.