• Anthropic, OpenAI and Elon Musk have backed stronger safeguards around increasingly powerful frontier AI
• Banks continue to accelerate AI adoption while placing greater emphasis on governance, validation and human accountability
• Cybersecurity concerns are rising as frontier models become capable of more sophisticated autonomous attacks
• Financial regulators are favouring controlled experimentation and existing accountability frameworks rather than blocking innovation
• The emerging competitive challenge for banks is demonstrating that greater AI capability does not mean diminished institutional control

The extraordinary intervention by some of artificial intelligence’s most powerful figures is strengthening an argument already familiar inside banking – that increasingly capable AI cannot be treated simply as another productivity technology.

Anthropic chief executive Dario Amodei has called for the pace of frontier AI development to be moderated, proposing independent safety evaluation, greater coordination between leading developers and international cooperation.

OpenAI chief executive Sam Altman and Elon Musk subsequently backed the broad call for stronger safeguards, while OpenAI has separately advocated mandatory, capability-based national safety requirements in the US.

For banks, the warnings arrive at a striking moment. Financial institutions are moving rapidly from AI experimentation towards deployment, while simultaneously demanding stronger governance around the technology.

OpenAI last week launched ChatGPT for Financial Services, developed with input from Morgan Stanley and Evercore, while Anthropic has now expanded Claude further into financial advice and wealth management.

That apparent contradiction – rapid adoption alongside demands for stronger safeguards – increasingly defines the banking sector’s response.

Banks have generally avoided calls to stop AI development. Instead, the emphasis has been on controlled deployment, traceability, human accountability and applying established risk disciplines to new models.

The American Bankers Association notes that generative and agentic AI remain emerging technologies for banks and stresses the importance of effective risk management, governance and controls.

Cybersecurity is adding urgency. The Bank for International Settlements’ Financial Stability Institute warned this month that frontier AI could autonomously identify vulnerabilities, develop exploits and conduct increasingly sophisticated cyber operations.

For financial institutions, it identified shorter remediation windows, greater breach risk and increased third-party dependencies among the potential consequences.

The concern therefore extends beyond whether a bank’s own AI produces an incorrect answer.

Increasingly powerful models could change the threat environment surrounding the entire financial system, while autonomous agents capable of taking actions create difficult questions over authorization, accountability and operational control.

UK regulators are pursuing a similarly pragmatic balance.

The Financial Conduct Authority has said it does not intend to create an entirely separate regulatory regime for AI, instead relying on existing requirements including Consumer Duty, governance expectations and the Senior Managers and Certification Regime.

At the same time, its Supercharged Sandbox is allowing firms to test advanced AI in controlled conditions, including agent-led payments, fraud detection, compliance and governance applications.

The result is a financial sector that appears neither anti-AI nor comfortable with unrestricted acceleration.

Banks are increasingly embedding AI into research, modelling, compliance, customer service and operational processes, but doing so within structures designed around validation, auditability and accountable decision-making.

The intervention from Amodei, Altman and Musk may therefore reinforce rather than disrupt banking’s direction of travel.

If the companies building the most powerful AI systems are themselves arguing that technological capability is advancing faster than existing safeguards, banks have another reason to resist a simple race towards maximum automation.

For financial services, the competitive advantage may ultimately belong not to institutions deploying the most AI, but to those capable of proving they remain in control of it.