Scoop: BNY gets access to OpenAI, Anthropic's advanced cyber models

BNY Mellon has obtained early access to advanced cybersecurity AI models from both OpenAI and Anthropic, positioning it among a limited group of vetted enterprises with such access. The bank, which manages $47 trillion in assets as the world's largest custodian bank, will use OpenAI's GPT-5.4-Cyber and Anthropic's Mythos models. CEO Robin Vince stated that the AI laboratories recognize the importance of releasing cyber-capable models to certain institutions early for feedback and real-world testing purposes.

OpenAI's GPT-5.4-Cyber is rolling out more broadly than Anthropic's Mythos model, according to the report. The arrangement allows BNY Mellon to test these advanced AI cybersecurity systems within its financial infrastructure. The early access program is designed to gather institutional feedback before wider release of the models.

This development occurs amid ongoing deployment of advanced AI models for cybersecurity applications by major AI laboratories. The access provides BNY Mellon with cutting-edge AI tools for potential threat detection and vulnerability assessment within financial systems. The bank's participation contributes to real-world testing of these models in high-stakes financial environments.