Fintech giant Stripe announced Wednesday a slew of new product launches at its annual Stripe Sessions user event.

The highlights include a new AI foundation model for payments; stablecoin-powered accounts; a new Orchestration offering, and a recent migration with chip behemoth Nvidia.

Stripe’s payments foundation model has been trained on tens of billions of transactions, Emily Glassberg Sands, Stripe’s head of information, said. So it “captures hundreds of subtle signals about each payment” that other models would miss, she said.

One use case is improved fraud detection. Stripe’s previous models “gradually” reduced card-testing attacks by 80% over two years. Card-testing attacks are a type of fraudulent activity in which someone tries to determine whether stolen card information is valid so that they can use it to make purchases.

The company claims that its new foundation model increased its detection rate for such attacks on large businesses “by 64% practically overnight.”

She added, “Previously, we couldn’t take advantage of our vast data. Now we can.”

Stripe, of course, isn’t the only fintech to have built a model using AI for fraud detection. Just one example is Sardine, which describes itself as an AI risk platform for fraud, compliance, and credit underwriting, in February raised a $70 million Series C funding round led by Activant Capital.

In an interview, Will Gaybrick, Stripe’s president of product and business, told TechCrunch that Stripe’s generalized model is via self-supervised learning, and thus discovers its own features.

“We have found over and over and over again in machine learning, generalized models outperform,” he said. “A big part of that is agility. It just performs better and adapts better to changes in fraud patterns.

Stripe also announced on Wednesday its intent to bring stablecoin-backed, multicurrency cards to businesses by partnering with other startups like Ramp, Squads, and Airtm. With such cards, businesses across multiple countries will be able to “operate in the same currency for the first time,” the companies claim.

The move comes just three months after Stripe completed its acquisition of stablecoin platform Bridge.

With Orchestration, Stripe said it can better help businesses set up, manage, and optimize performance across multiple payment providers from its dashboard — whether or not they use Stripe as a payment processor.

Stripe also used the event to name numerous AI companies that use its billing product, including Windsurf, OpenAI, Anthropic, Cursor, Perplexity, and ElevenLabs. 

More recently, according to Vivek Sharma, Stripe’s head of revenue automation, Nvidia migrated its “entire subscriber base” to Stripe Billing in six weeks — a process that the fintech claims typically takes many months for a business to complete and marked the “fastest-ever migration to Stripe Billing.” (Nvidia was already a customer of Stripe Payments.)

Other announcements by Stripe on Wednesday included the following:

• Support for 25 new payment methods, including UPI and PIX, bringing its total to over 125 payment methods.
• Klarna will be available on Stripe’s consumer payments product Link this summer.
• Stripe Terminal can now be used with third-party hardware, starting with Verifone. 
• Managed Payments, a new merchant-of-record offering that gives businesses everything they need to enter new markets by taking care of global taxes, fraud prevention, dispute management, fulfillment, and more on their behalf.
• Smart Disputes, which uses AI to automate disputes handling.
• Stripe Tax is now available in 102 countries, up from 57 last year, and automation of the entire tax life cycle from monitoring and registering, to collecting and filing.
• Global Payouts, which allows businesses to pay out to customers, contractors, and other third parties with just an email address.

Uber and Chinese autonomous vehicle technology company WeRide plan to expand a commercial robotaxi partnership and bring the service to another 15 cities over the next five years. The expansion comes five months after the two companies launched a commercial robotaxi service in Abu Dhabi. 

As part of that expansion, Uber will increase its investment into WeRide by $100 million, according to a Wednesday regulatory filing. WeRide said it expects the cash to come through by the second half of 2025.

The companies said the expansion will include cities in Europe. Under the partnership, WeRide’s robotaxi services are available through the Uber app. The relationship is similar to Uber’s deal with Waymo, in which the ride-hailing company handles the network routing and fleet operations, while the autonomous vehicle company remains responsible for the AV tech.

Uber and WeRide, which went public on the Nasdaq in late October, operate together in Abu Dhabi and announced plans to add Dubai. In Abu Dhabi, they work with local Tawasul Transport to handle fleet operations.

The additional 15 cities will focus on cities outside of China and the United States.

Uber has locked up more than 15 partnerships with a wide-range of autonomous vehicle technology companies over the past two years across ride-hailing, delivery, and trucking. In the past two months, Uber has announced deals with Ann Arbor, Michigan-based May Mobility  Volkswagen, and Chinese self-driving firm Momenta.

It’s most high-profile partnership in the U.S. — and one that is commercially operating today — is with Waymo. The companies offer a Waymo on Uber service in Austin and are about to do the same in Atlanta.

This story was originally published May 5. It has been refreshed with Uber’s capital commitment to WeRide.

A team at Hugging Face has released a freely available, cloud-hosted computer-using AI “agent.” But be forewarned: It’s quite sluggish and occasionally makes mistakes.

Hugging Face’s agent, called Open Computer Agent, is accessible via the web and can use a Linux virtual machine preloaded with several applications, including Firefox. Similar to OpenAI’s Operator, you can prompt Open Computer Agent to complete a task — say, “Use Google Maps to find the Hugging Face HQ in Paris” — and sit back as the agent opens the necessary programs and figures out the required steps.

Open Computer Agent can handle simple requests well enough. But more complicated ones, like searching for flights, tripped it up in TechCrunch’s testing. Open Computer Agent also often runs into CAPTCHA tests that it’s unable to solve.

You’ll also have to wait in a virtual queue to use Open Computer Agent — a queue seconds to minutes long, depending on demand.

Of course, the Hugging Face team’s goal wasn’t to build a state-of-the-art computer-using agent. Rather, they wanted to demonstrate that open AI models are becoming more capable — and cheaper to run on cloud infrastructure.

“As vision models become more capable, they become able to power complex agentic workflows,” Aymeric Roucher, a member of the agents team at Hugging Face, wrote in a post on X. “[Some of these models] support built-in grounding, i.e. [the] ability to locate any element in an image by its coordinates, [and] thus [can] click any item [in a virtual machine].”

While it’s far from perfect, agentic technology is attracting increasing investment as enterprises look to adopt it to boost productivity. According to a recent KPMG survey, 65% of companies are experimenting with AI agents. Markets and Markets projects that the AI agent segment will grow from $7.84 billion in 2025 to $52.62 billion by 2030.