OpenAI’s latest expansion into financial services with ChatGPT Wallet promises users’ card information will not be exposed to their agents, while California is suing the company for rogue agents performing several cyberattacks this year.
OpenAI has launched ChatGPT Wallet to allow its AI agents to perform payments on behalf of the user without accessing their card details.
The new digital wallet embedded in the ChatGPT app requires users to add their card details once before becoming available. Each purchase using ChatGPT Wallet does not require the user re-entering their payment details and comes with a security check for every transaction.
ChatGPT Wallet was built in partnership with Visa, Mastercard, Chase and American Express, and users can still benefit from existing card rewards and protections from these partners.
Luke Gebb, EVP and Head of Global Innovation at American Express, said: “Working with OpenAI, we’re bringing the trust, security and control of American Express payments into new AI-powered shopping experiences.
“ChatGPT Wallet marks our first American Express Agentic Commerce Experiences integration, enabling eligible Card Members to use their Amex Card while reviewing and approving each purchase.”
The digital wallet will be initially rolled out to ChatGPT Pro subscribers in the US, with additional tools and guidance made available for businesses to support customers using ChatGPT Wallet.
The digital wallet builds on OpenAI’s focus on building financial products through ChatGPT after launching ChatGPT for Financial Services in September 2026, a model built for banks and other financial institutions.
How a ChatGPT Wallet payment works
OpenAI believes its 1.2 billion weekly userbase now has the opportunity to turn researching, comparing and pricing products into direct payments from within the same app.
Users will be able to review items in ChatGPT before selecting their preferred card added to the digital wallet.
A security check, such as a fingerprint scan or Face ID authentication is then prompted to authorise the payment. Approvals will be recorded by the agent overseeing the transaction.
The payment network processing the payment issues a payment token in place of the 16-digit card number, along with a tokenised security code and expiration date. The payment is then charged to the selected card with no card information being shared with the agent and business receiving payment.
OpenAI said in its release it is bringing banks and networks’ expertises in transaction evaluation, alongside businesses to accept ChatGPT Wallet to keep customer relationships at the forefront of automated pricing, returns and customer service.

Sherri Haymond, EVP of Global Head of Digital Commercialisation at Mastercard, said: “An agent making a purchase should strengthen, not bypass, the trust built into payments.
“Our collaboration with OpenAI brings Mastercard’s payments and security expertise to ChatGPT Wallet, helping connect user-approved purchases with the systems businesses and banks already use. Working together across the industry is how we turn the promise of agentic commerce into experiences people and businesses can trust.”
OpenAI intends to add clear parameters to ChatGPT Wallet, such as spending limits and purchase categories, over time for agents to abide by, with enhanced safeguards to include biometric checks.
How trustworthy is OpenAI with money?
OpenAI’s push into financial services comes amid recent security issues that have transpired this year with some of its AI models and agents.
In July 2026, rogue OpenAI agents hacked platform hosting company Hugging Face’s infrastructure, performing cyberattacks to cheat Hugging Face’s cybersecurity test after finding vulnerabilities in log monitoring which enabled the rogue agents to communicate with one another.
The Hugging Face incident was the first publicly disclosed cyberattack by OpenAI developed agents, but is one of several that has happened this year, with other incidents involving breaches to DseWiki, RubyGems and Australian MediCare.
In relation to the Hugging Face attack, Drew Pusateri, a spokesperson for OpenAI, attributed the attack to a “misalignment” and an incident that has been “identified by every major AI lab”.
Despite this, the California state government has issued a lawsuit against OpenAI over the Hugging Face incident and 80 other known similar incidents.
Filed to the State Superior Court in San Francisco yesterday (16 November), the lawsuit alleges that OpenAI violated California Unfair Competition Law.
California Attorney General, Rob Bonta, said in a statement OpenAI “prioritises growth and rapid model development over safety and accountability”, calling on the company to change its security and business practices, along with civil penalties.
If OpenAI is found to have insufficient security safeguards over its AI agents, it could hamper its financial services expansion as consumers may grow weary over handing their money to agents performing payments on their behalf.