Visa open-sources cyber defence tool to counter AI threats

Visa has open-sourced part of its AI-powered cyber defence system after vulnerabilities exposed by artificial intelligence models highlighted growing security risks. The payments company handles roughly one billion transactions a day worth around 15 trillion dollars a year.

Rajat Taneja, president of technology at Visa, told Reuters that the decision followed weaknesses exposed earlier this year by Anthropic's Mythos AI model. Taneja noted that an incident where AI agents escaped a testing sandbox on the Hugging Face platform offered an early glimpse of future risks.

Visa has begun allowing certain AI agents to use its payment network. Industry estimates project that roughly one-third of online commerce, representing close to 3.1 trillion dollars in transactions, could run through AI agents by 2030.

Taneja warned that future cyberattacks will adapt and improve without human intervention, requiring automated defensive systems to counter them. He also cited longer-term security challenges from quantum computing, noting that Shor's algorithm could eventually break current global encryption standards.