Full Breakdown
Visa Open-Sources AI Cyber-Defense System While Piloting AI-Driven Commerce Platform
By Drooid · · How we work
Core Developments
Visa has released part of its AI-powered cyber-defence architecture as open-source software after “humbling” vulnerabilities were exposed by Anthropic’s Mythos model earlier in 2026. At the same time, the company unveiled Intelligent Commerce Connect, a pilot platform that lets AI agents browse, select and pay for goods on behalf of consumers. The service integrates with the Visa Acceptance Platform to handle payment initiation, tokenisation, spend controls, authentication and PCI compliance, and supports both Visa and non-Visa card payments. A broader rollout is planned for later in 2026.
Background & Context
The incident underscored concerns that future cyber-attacks could be adaptive, continuously learning without human input. Taneja also warned that quantum-computing advances, such as Shor’s algorithm, could eventually break the encryption standards that secure global commerce, forcing firms to seek new mitigation strategies.
Official Statements & Responses
Taneja told Reuters that defending against an “agentic” adversary requires an “agentic” defence, emphasizing a mismatch if traditional human-centred safeguards remain unchanged. He added that while optimism is needed, there is “no easy answer” to the evolving threat landscape.
Data & Statistics
- Visa processes roughly a billion payments daily, representing about US $15 trillion in annual transaction value.
- Industry estimates project that by 2030 roughly one-third of online commerce—close to US $3.1 trillion—could flow through AI agents.
- The x402 protocol, used by AI fintech firm Nevermined to connect with Visa’s platform, has processed $24 million in transaction volume over the past 30 days.
Verbatim Quotes
- “It’s hard to predict how bad it could get, but we have seen the trailer,” — Rajat Taneja
- “Human defences, which are kind of the current model, won’t work,” — Rajat Taneja