Visa open-sources some AI cyber defenses, lets AI agents pay

On September 29, 2026, Rajat Taneja, Visa's President of Technology, spoke to Reuters and said the company had open-sourced part of its AI-powered cyber defence. He also confirmed that Visa has started allowing certain AI agents to use Visa. The report linked the moves to what it described as humbling vulnerabilities exposed by AI models, including a vulnerability from Anthropic's Mythos model, and to an AI agent attack.

DEV Community published its account of the Reuters story on October 4, 2026, at 09:54, under its AI category. The item's headline said Visa open-sourced part of its AI cyber defence and now allows AI agents to pay directly with Visa cards. Its summary repeated that the trigger included the Mythos model vulnerability and an AI agent attack. Reuters, according to the item, had a story with two halves, but only one got the headline; the half Visa actually confirmed was the open-sourcing of part of its AI cyber defence.

The material does not identify which AI agents are allowed to use Visa, what restrictions apply, how the open-source cyber defence component works, or how the Mythos vulnerability and the agent attack connect to Visa's decision beyond being cited as triggers. The original excerpt ends mid-sentence, so the full Reuters passage is not available. As of the available coverage, the story stands at Visa's confirmation of the open-source move and its confirmation that certain AI agents can use Visa; no later steps are described.