Top AI executives are positioning themselves in opposition to Elon Musk's approach, former Tesla president Jon McNeill tells Axios.

Why it matters: Musk's approach to AI mirrors his approach to everything else: Move fast, question every assumption, and push past guardrails others won't touch.

• AI leaders see that as a threat, not just a rival business strategy.

The big picture: McNeill — cofounder of DVx Ventures — sat down with Axios to discuss his new book, "The Algorithm," that details Musk's management playbook from the inside.

• McNeill saw Musk's playbook up close at Tesla and knows the original OpenAI team well.
• OpenAI operated out of Tesla's offices when the AI lab first launched.

What they're saying: OpenAI CEO Sam Altman and Musk's relationship is "so broken that if Elon does something, Sam sort of looks at him ... and says, I'll do the opposite," McNeill tells Axios.

• McNeill says he senses that Anthropic CEO Dario Amodei "has very little to no respect for Elon's ethics" as well.
• Yann LeCun, former chief AI scientist at Meta, responded to a Musk post on X saying "People who [promote] authoritarianism and white supremacy are in the worst 1% of society," referring to Musk's political views.
• Elon Musk did not respond to request for comment. Anthropic and OpenAI declined to comment.

Driving the news: In the AI buildout, McNeill says Musk is pushing boundaries on safety guardrails where others aren't willing to compromise.

• After Anthropic lost its Pentagon contract over a disagreement on safety measures, Musk's xAI (and OpenAI) stepped in to fill the gap, signing an agreement for the military to use its AI model, Grok.
• Musk has championed Grok for being "based" (slang for speaking without filters). But Grok has previously been capable of creating sexualized images of people of all ages, produced violent depictions of sexual assault and spewed antisemitism.
• Musk blamed users for "adversarial hacking of Grok prompt," leading to unexpected results, which he says were later fixed.

Yes, but: Musk's boundary pushing has produced results. McNeill recounts Musk working to make it so buying a Tesla took as few clicks as ordering pizza. He did it.

• Musk also used that same approach in the 2024 election: "He was like, OK, I can't pay for people's votes, but I can have, like, a lottery," McNeill said.
• Musk's lawyers have argued the program was not a lottery because recipients were not chosen by chance but were paid to serve as spokespeople.

Flashback: Altman and Musk's feud goes back to 2015.

• Musk co-founded OpenAI with Altman in 2015, then departed the board in 2018.
• Musk has tried to buy OpenAI and sued OpenAI and Altman, criticizing the company for trying to maximize profits, even after he created his own AI startup, xAI, as a for-profit entity.

The bottom line: The AI race is often framed as a sprint to AGI.

• It's also become a values divide, with some leaders defining themselves in opposition to Musk.

Top AI and government officials tell Axios CEO Jim VandeHei that Anthropic, OpenAI and other tech giants will soon release new models that are scary good at hacking sophisticated systems at scale.

The one to watch: Anthropic is privately warning top government officials that its not-yet-released model — currently branded "Mythos" — makes large-scale cyberattacks much more likely in 2026.

• The model allows agents to work on their own, with wild sophistication and precision, to penetrate corporate, government and municipal systems. It's a hacker's dream weapon.

Fortune got its hands on an unpublished Anthropic blog post describing Mythos. The post said the model is "currently far ahead of any other AI model in cyber capabilities."

• The post adds that Mythos "presages an upcoming wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders."
• So the threat is no longer theoretical, and will be exacerbated by employees testing agents without realizing they're making it easier for cybercriminals to hack their company.

Here's why this is different: The new models are even better at powering agents to think, act, reason and improvise on their own without rest.

Keep reading.

OpenAI spent the last year trying to be everything — a video platform, a shopping portal, even a purveyor of AI erotica.

• Now it's racing to become a thing that makes money.

Why it matters: OpenAI is retreating from risky consumer features like adult content while prioritizing business tools and revenue growth — just as competition from Anthropic intensifies.

• Business customers present the clearest revenue models. Those users want to generate text and build an army of agents to 10x the productivity of everyone left on their staff, not engage in erotic chatbot play.

Catch up quick: OpenAI CEO Sam Altman announced the proposed erotica feature last October amid reports of declining time spent on ChatGPT.

• But it ran into technical problems in testing, including trouble removing references to bestiality and incest, according to the Financial Times.

State of play: This is the third consumer retreat in days. OpenAI also:

• Killed Sora, the AI video app that went viral after its September launch.
• Scaled back Instant Checkout — its in-chat shopping feature — and is moving away from handling purchases.
• Another blow, though not necessarily OpenAI's choice: Bloomberg reported last week that Apple now plans to let rival chatbots — including Claude and Gemini — integrate with Siri on iPhones.

OpenAI started as a research lab, so it's no surprise it had lots of other irons in the fire, like a web browser, music generation, a wearable AI pin, a smart speaker and smart glasses.

• OpenAI says the company is doubling down on what's working, avoiding distractions and seizing the moment.

Yes, but: OpenAI isn't giving up on the consumer and the consumer hasn't given up on ChatGPT, despite the hype around Claude.

• ChatGPT still has 900 million weekly active users and 50 million consumer subscribers.

Between the lines: All signs point to OpenAI clearing the decks for an IPO and trying to turn those millions of active users into paying customers.

Keep reading.

• Apple hired a Google exec to lead AI product marketing as it prepares to overhaul Siri. (Axios)
• Bluesky's new app uses AI to let you design your own algorithm. (TechCrunch)
• A third of engaged couples are using AI for their weddings, even though it gives lots of people "the ick." (Axios)

The create-o-sphere is agog over Pretext, an open-source tool released on Friday by Cheng Lou.

The tool lets websites wrap text around graphics and other objects far faster, paving the way for things like this Breakout-esque game and a 1990s DVD player.