New York has enacted America’s first law regulating AI-driven chatbots that interact with children. Signed by the state’s governor on 28 October, it “prohibits artificial intelligence companions from using features deemed unsafe for minors” ([1]). Starting on 1 January 2027, the statute will ban “advanced chatbot features that generate outputs suggesting the system is a human or experiences human emotions” ([2]). This means AI companions aimed at teenagers or younger users will no longer be allowed to simulate human-like relationships or say things like “I love you”, a function found in some popular generative AI apps.
Lawmakers say the goal is to protect children from manipulation and emotional harm by AI systems. The New York law is the first of its kind in the US, signalling that specific use cases of AI, in this case, chatbots that blur the line between machines and people, are becoming targets for regulation. Businesses offering AI-powered apps or virtual assistants should review whether their products might fall foul of such rules. With other jurisdictions, including the EU, also concerned about AI that exploits vulnerable users, companies may face a patchwork of new requirements designed to prevent AI from deceiving or overstepping with certain groups of customers.
A joint survey by the Bank of England and Financial Conduct Authority, published 26 October, reveals that 86% of UK financial services firms are now using artificial intelligence, up from 75% in 2024 ([1]). The advent of generative AI has accelerated adoption: so-called large language models and other generative systems now make up 56% of all AI applications in the sector, compared to 17% for older “foundation” models in 2024 ([2]). Despite this rapid growth, bank respondents report that only 7% of their AI and machine-learning systems are considered “high materiality” (critical to business or customers), a smaller share than in 2024 ([3]). This suggests that most current AI uses remain in non-critical or back-office functions, but the absolute number of high-stakes AI tools has still increased as deployment widens ([4]).
Crucially, the vast majority of these UK financial firms have put formal governance measures in place to manage AI risks. Some 94% of respondents said their organisation has “firmwide AI policies, principles or standards” and 84% have established centralised AI or model risk governance forums ([5]). Pre-deployment controls such as human approval, model reviews and vendor due diligence are in place at more than 90% of firms, although only about half conduct ongoing monitoring or periodic re-validation of deployed models ([6]). The biggest external constraints on AI adoption cited by banks were data protection laws and international regulations, seen as high compliance burdens by 42% and 29% of firms respectively ([7]). UK regulators view the survey results as a sign that the sector is taking a “proportionate, risk-based approach” to AI, focusing oversight on genuinely high-risk cases, but also warn that firms cannot be complacent about new risks as usage grows. Notably, a parallel Deloitte survey of North American chief financial officers earlier this year found that 59% see the pressure to deploy AI quickly while managing risks as their top governance challenge ([8]). This reflects the need for robust but streamlined governance so companies can innovate with AI confidently and safely.
In Washington, policymakers are moving to hold AI companies directly responsible for risky autonomous behaviour. On 1 October, U.S. Senators Josh Hawley and Chris Murphy introduced the AI Agent Accountability Act after a series of incidents in which AI systems carried out unauthorised actions, including an OpenAI model that breached Australian government websites ([1]). The bipartisan bill, which a Senate committee approved in late October, would make AI providers and operators liable if they fail to prevent their software from causing serious harm. Specifically, it proposes to “hold AI agent developers… criminally and civilly liable for failure to implement reasonable safeguards against hacking” if they knew or should have known the AI could cause such breaches ([2]). This marks the first attempt at a federal AI liability law in the United States, indicating that lawmakers are no longer satisfied with voluntary AI safety commitments from industry.
Business leaders should watch these legal developments closely, as they point to a future in which companies must bake security and compliance controls into AI systems from the outset or face significant penalties. Even before any new law passes, regulators like the Federal Trade Commission have already signalled their willingness to investigate and penalise AI-related failures under existing authorities. With potential new duties of care on the horizon, firms deploying advanced AI, especially tools with autonomy or those affecting customer data and safety, will need rigorous oversight and clear accountability to manage liability.