BNP Paribas has blocked employee access to Anthropic's Claude AI models across its Asia operations, the bank confirmed this week. The decision, first reported by Crypto Briefing, underscores growing regulatory and security challenges that global banks face when deploying third-party AI tools.
Why BNP Paribas blocked Claude
The restrictions apply to Claude models used by staff in Asian offices. The move isn't a blanket ban on all AI—it signals a specific concern about compliance and data security with external large language models. Financial institutions in Asia have been under increasing pressure from local regulators to vet AI providers for data residency, model transparency, and potential cross-border data flows.
BNP Paribas hasn't detailed which regulators or specific rules triggered the block. But the action fits a broader pattern: banks are rethinking whether off-the-shelf AI from U.S. firms meets the stricter governance standards now emerging in markets like Singapore, Hong Kong, and Japan.
Shift toward in-house AI
The Claude block is prompting BNP Paribas to accelerate its own internal AI development. Rather than rely solely on external models, the bank is pushing resources into proprietary solutions that can be tightly controlled and audited. That's a trend visible across the industry—JPMorgan, Goldman Sachs, and others have similarly invested in private AI infrastructure this year.
For BNP Paribas, going in-house means longer development timelines but fewer compliance headaches. The trade-off is real: internal models often lack the raw performance of frontier systems like Claude, but they offer banks full visibility into training data, outputs, and security protocols.
What this means for banking AI
BNP Paribas's decision isn't an isolated incident. Regulators in Asia have been especially active on AI governance. In April, Singapore's Monetary Authority released updated guidelines on responsible AI use in finance, emphasizing the need for explainability and human oversight. Japan's Financial Services Agency followed with a similar consultation paper in May.
The timing isn't great for Anthropic, which has been pitching Claude to enterprise clients as a secure, compliant alternative to ChatGPT. Banking is a key vertical for the company's growth. Losing a customer like BNP Paribas—even partially—could push other financial firms to reconsider their own reliance on third-party AI.
What happens next? BNP Paribas is expected to roll out its first in-house AI tools for Asian staff by the end of the third quarter. Whether those tools can match Claude's capabilities—and whether other banks follow BNP's lead—remains an open question.
