AI is outpacing guardrails. Businesses must build their own

Dario Amodei, CEO of Anthropic, joined Silicon Valley's concerns about AI development speed in a recent letter. He warned of a potential AI-driven internet takeover causing billions in damage within six to twelve months. However, we don't need to wait that long to see AI's harm without proper governance. Most immediate damage comes from ordinary systems, not rogue agents.

For instance, AI is already used in banking to decide loan eligibility, flag claims, and close accounts. Deloitte found that AI usage in banking doubled within a year, and AI incidents increased eightfold from 2022 to 2025. While businesses can control their AI deployments, regulators are slow to act. To address this gap, businesses must develop their own AI governance framework.

This framework should include three key steps: identifying an owner for each AI system, maintaining human oversight, and incorporating vendors into governance strategies. Naming an owner for each AI system ensures accountability and quick action when needed. Employees should receive training to evaluate AI outputs, override decisions if necessary, and explain outcomes clearly to customers and regulators.

Including vendors in governance frameworks is crucial as businesses often buy AI rather than build it. This involves adding audit rights, bias-testing support, incident notification, and termination rights to vendor contracts. While the situation may seem dire, AI also offers benefits like reduced customer friction, increased employee engagement, and expanded economic opportunities.

Implementing proper AI governance can ensure these benefits are realized without compromising security or fairness.