FCA warns firms against purely technical explanations of AI decisions
Research says AI models may be easier to read than thought. The FCA wants explanations of AI decisions that show customer outcomes, not mechanics.
Topic
Research says AI models may be easier to read than thought. The FCA wants explanations of AI decisions that show customer outcomes, not mechanics.
FCA good practice examples describe firms choosing where AI supports, automates or steps aside. Uneven AI ability makes that line hard.
Six banks have published principles for AI agent payments. Who pays when an agent is tricked is still open, and insurance cover may not fit.
A poisoned release of an open source AI gateway shows how one software supply chain compromise could reach many cyber policyholders at once.
Royal Bank of Canada told staff that such agents can create liability exposure, according to a memo seen by Bloomberg.
The vendor metric covers all Anthropic's paying customers, not insurers, but it suggests why AI budgets set on a pilot may understate the bill.
The London start-up has raised $2m for software that sets what AI agents may ask, and what customers must see, before an insurance quote.
Allianz Re has also agreed to enter a multi-year reinsurance agreement with the US firm, which writes only through wholesale brokers.
Insurers run a median of 18 AI systems in production, against 22 at UK banks and 43 at international banks.
Its US share sale filing says the market for AI-specific cover is nascent and may not offer terms at all.
Its examples also show firms keeping a central inventory of AI agents and avoiding dependence on a single model provider.
The US comparison site has named OpenAI's Dots, Meta's Muse and Instinct among the personal agents with access to its quotes.
The US exposure management firm's portfolio agent answers plain-language questions about an insurer's whole book and runs the analysis on screen.
Verisk has also added premium audit resources and a hail and wind history connector to its offering inside Anthropic's Claude.
Chief claims officer Stephen Deane will also take on operations when the changes take effect on 2 November.
Rushil Goel's San Francisco group starts with commercial insurer Nirvana, claims administrator Vega and operations platform Cooper.
The assistant can plan, build and test life insurance models from a specification, then document the work for actuaries to review.
Only 41% of the 58 insurers, reinsurers and brokers surveyed consistently track the benefits of each AI use case.
The home insurance distributor has also launched a service that checks cover against seven insurers several times a year.
The US carrier's quoting agent Aimee prices small business risks from a plain-language chat, with agents reviewing quotes before binding.