Making companies pay for harms caused by AI could be one of the strongest incentives for preventing those harms in the first place.

Why it matters: Liability creates a direct financial incentive to identify risks, test for them and build protections before AI systems are deployed, not after someone is harmed.

The big picture: Governments have several ways to make companies bear the cost of AI harms — from agencies imposing regulatory fines to courts awarding civil damages.

• With federal AI legislation moving slowly, people harmed by AI are increasingly turning to existing laws and the courts.

Driving the news: The increased use of agents has just made the liability questions much more urgent.

• Agentic systems are getting more access to websites, accounts and other real-world systems.
• In Australia, a user asked his AI agent to get into a sold-out fitness class and the agent allegedly hacked the booking system and bumped another person from the class.
• The stakes were low — but the episode offers a glimpse of a complicated future where agents gain even more access.

Reality check: Liability is not a yes-or-no question.

• A core dispute is whether companies should be strictly liable for harm caused by their agents, or only when they failed to take reasonable precautions. And responsibility can be divided among a model maker, a cloud provider, a company that customized and deployed an agent, and the user who tasked it.
• Open-source models make that allocation even thornier. If a business builds an agent on an open model hosted in the cloud, and that agent harms another company, which actor should pay?

Zoom out: Governments are already using several different legal tools to make AI companies bear the costs of harms their tools create.

• The EU has begun enforcing its AI Act, which allows regulators to fine companies that fail to disclose certain AI interactions or appropriately label AI-generated content. More consequential rules for high-risk systems are still ahead.
• In December, a separate EU product-liability directive will make it easier for people to seek compensation for harm caused by defective commercial AI software.
• The Take It Down Act, the main federal AI-specific law enacted so far, criminalizes publication of nonconsensual intimate imagery, including AI-generated imagery, and requires covered platforms to remove validly reported material within 48 hours.

Meanwhile, actions in court are moving forward.

• OpenAI is the subject of a dozen lawsuits alleging ChatGPT contributed to wrongful death, mental distress and "dangerous public nuisance."
• Patients have sued health insurance companies claiming that AI systems improperly denied doctor-recommended care.
• Ryan Clarkson, whose firm has brought several lawsuits against AI providers, argues existing consumer-protection and product-liability law can do much of the work even as governments debate new AI rules.
• "There is complete political gridlock in Washington," Clarkson said, suggesting that firms like his can act almost like "private attorneys general."

Yes, but: University of Washington law professor Ryan Calo tells Axios courts are unlikely under current law to impose strict liability on AI makers — at least if they see AI as socially useful.

• That would leave plaintiffs to show a company was negligent in its testing, release, monitoring or safeguards.

Between the lines: AI's opacity can make cases harder to prove, but it doesn't make liability impossible.

Keep reading.

Elon Musk and Mark Zuckerberg have muscled their way back into AI's elite ranks, defying early obituaries to close the gap on a new generation of Silicon Valley startups.

Why it matters: Recent gains by tech giants SpaceX and Meta — long stuck in AI's second tier — are putting new pressure on a hierarchy dominated by OpenAI and Anthropic.

State of play: Both companies released models this week featuring a combination of performance and price that would have been hard to imagine from either lab a year ago.

• SpaceX's Grok 4.6 scored essentially even with OpenAI's GPT-5.6 Sol Max on the closely watched Artificial Analysis Intelligence Index, and just behind Anthropic's Fable 5 Max.
• Promising "something special," Musk said Grok 4.7 should arrive in three to four weeks and predicted it will "exceed all current models" after additional training on a massive trove of SpaceX data.

Meta, meanwhile, is trying to squeeze the frontier from below.

• Its new models deliver competitive performance at dramatically lower cost — including Muse Glimmer, an open-weight system small enough to run locally on a laptop.
• Zuckerberg cast the push as "superintelligence for everyone" in a sweeping manifesto this week, reviving Meta's open-AI ambitions as the U.S. races to regain ground from China.

Catch up quick: As OpenAI and Anthropic pulled ahead, Musk and Zuckerberg answered with the full force of their empires — spending billions on talent, infrastructure and acquisitions.

Keep reading.

A new "digital organism" could simulate how a potential drug or genetic change moves through molecules and cells and eventually people, according to a new Nature Medicine paper out today.

Why it matters: Better AI simulations could streamline biology and help researchers screen ideas earlier and focus experiments on the most promising candidates to catch risks sooner.

The big picture: The scientists describe the project as a kind of flight simulator for medicine. Scientists will be able to run experiments on biology, like altering a gene or introducing a new drug to a cell, before attempting those experiments in the real world.

How it works: Existing biology AI generally handles one specialized task, such as predicting a protein's shape. In the Nature article, Eric Xing and his co-authors argue that the next step is connecting those systems — models for DNA, proteins, cells, tissues and health outcomes — into a dynamic simulator.

• "It is really about building a simulator of biology," Xing told Axios.

Keep reading.

• Google unveiled its latest crop of Pixel devices, including new AI-powered camera and health features. (Axios)
• Our colleague Joann Muller has the scoop on how Amazon-owned Zoox decided its robotaxis were ready for prime time. (Axios)
• Apple has hired former American Airlines executive Nate Gatten to head its government affairs efforts. (Bloomberg)

Mady here, dying to talk to the investor who told the Financial Times that Anthropic may deserve a $3 trillion valuation.

• An investor said: "If Anthropic is growing 800 per cent a year, you'd think at the incredibly low end they would trade at 30 times [revenue]," and "That would make them a $3tn company."
• That would be about triple the valuation of the largest IPO in history, SpaceX.