By James W. Moore
The Industry’s New AI Leaderboard Ranks Allianz First. Read the Fine Print Before You Circulate It.
Carrier Management this week published a detailed breakdown of the second annual Evident AI Index for Insurance, a benchmarking exercise that ranks 30 large insurers and reinsurers on AI maturity across four pillars: Talent (weighted 45%), Innovation (30%), Leadership (15%), and Transparency (10%). Allianz took the top spot, leapfrogging last year’s leader AXA. Liberty Mutual ranked as the top pure P&C carrier at fifth overall, Zurich was the biggest mover, jumping from 12th to 4th, and State Farm posted the largest decline, falling to 24th.
The workforce numbers are the substance. By Evident’s count, one in every 50 employees across the 30 insurers is now an AI specialist, and AI-specialist roles grew 32% over the past year even as overall insurer employment declined slightly. Eleven of the 30 companies now have a Chief AI Officer or equivalent, and two-thirds of those executives have held the role for less than a year. On the deployment side, 20 of the 30 insurers now document at least one AI use case with a reported outcome, up from 12 a year ago. But the outcomes tell their own story: 75% of documented results are productivity gains, while revenue uplift shows up in just 2% of them. Claims management leads deployment at 28% of use cases, and only 8% of use cases have advanced to the multi-agent, end-to-end work Evident sees as the 2027 frontier.
Now the fine print. Evident does not publish the underlying index values or pillar scores, only the rankings. Carrier Management emailed Evident asking about the data sources behind the talent counts and did not receive a response. The report’s own methodology description points to LLM extraction and machine learning tools run against public disclosures, press releases, and third-party data platforms. That is a defensible approach for directional analysis. It is not the same thing as audited data.
Why This Matters for Insurance:
This ranking will circulate through boardrooms this month, and executives at carriers that moved up will cite it while executives at carriers that fell will quietly ignore it. Both reactions miss the more useful signal. The disclosure gap is the real finding: nearly half of all well-documented AI use cases in the sector come from just five companies (Allianz, AXA, Manulife, Travelers, and Zurich). Everyone else is either doing less than they claim or disclosing less than they do, and an index built on public disclosures cannot tell those two apart. If your carrier’s ranking surprised you in either direction, the first question is not “how do we move up?” It is “how much of our actual AI work is visible to an outside analyst scraping our public statements?” For some carriers, a low rank is a strategy problem. For others, it is a communications artifact.
Sources:
- Ranking: Who Are the Insurance Industry’s AI Talent, Maturity Leaders? — Carrier Management
- Evident AI Index for Insurance — Evident Insights
The Bank of England Just Said the Quiet Part: Nobody’s Rulebook Was Written for AI Agents.
Speaking at the European Central Bank Forum on July 1, Bank of England Deputy Governor Sarah Breeden said the Bank is reviewing whether existing regulatory frameworks can cover agentic AI in finance, systems that make decisions and carry out tasks without direct human instruction. Her core admission deserves attention: current frameworks were not built to contemplate autonomous agents in payments, trading, and operations, and expecting a human to review every action these systems take is unlikely to be practical. This from a regulator that as recently as last year maintained existing rules were sufficient for AI risk.
The adoption data behind the concern comes from a 2026 Cambridge Centre for Alternative Finance report finding that 81% of surveyed financial services firms are adopting AI at some level, and 52% are already actively adopting agentic AI, mostly in internal functions like process automation and knowledge management. Breeden flagged two systemic worries. First, herding: autonomous agents trained on similar data or built around similar objectives could respond to the same market signals in the same way at the same time, amplifying volatility. Second, cyber resilience, where she described a step change in AI capability that cuts both ways, strengthening defenders while giving attackers tools that chain actions at scale and speed. She noted open-source models may trail the most capable closed models by only four to eight months, which limits the comfort regulators can take from export restrictions on frontier systems.
The remedies under discussion are blunt instruments: market-wide circuit breakers and kill switches to halt trading if faulty AI models drive severe disruption, stronger recovery requirements for core systems, and even arrangements allowing one bank to take over another’s basic functions during an outage. The Financial Stability Board, meanwhile, published a June consultation proposing 12 sound practices for responsible AI adoption covering governance, risk management, and third-party risk, explicitly flagging AI agents as a distinct challenge for human oversight.
Why This Matters for Insurance:
Every question Breeden raised about banking lands on insurance with the serial numbers filed off. Carriers are deploying agentic systems in claims right now, not hypothetically. The same herding logic applies: if a dozen carriers run agentic underwriting or claims triage tuned on similar data from the same handful of vendors, their models will make correlated mistakes, and correlated mistakes in a regulated industry become market conduct exams. The cyber framing matters twice for carriers, once as operators of these systems and once as underwriters of everyone else’s. And the FSB’s third-party risk emphasis should sound familiar: an insurer’s agentic AI almost always runs on infrastructure it does not own, governed by contracts its board may not have read. The supervisory conversation Breeden started will reach state insurance regulators. The carriers that fare best will be the ones that started answering these questions before being asked. The global regulatory picture, and why no two major jurisdictions are answering it the same way, is the subject of next Wednesday’s long-form article.
Sources:
- Bank of England Reviews AI Rules for Agentic AI in Finance — AI News
- FSB Consultation on Sound Practices for AI Adoption — Financial Stability Board
AI Failure Reporting Is Becoming Real Infrastructure, by Statute and by Standards.
Two developments this week, one from a statehouse and one from a security lab, point at the same emerging reality: the machinery for formally reporting AI failures is being built now.
On July 6, Illinois Governor JB Pritzker signed SB 315, the AI Safety Measures Act, making Illinois the first state to require independent third-party audits of large AI developers’ safety practices. The law applies to developers with more than $500 million in annual gross revenue and takes effect January 1, 2028. Covered companies must publish explanations of how their products could pose catastrophic risk, report critical safety incidents to the state within 72 hours, retain an annual independent compliance auditor with demonstrated frontier-model expertise, and maintain whistleblower protections. Fines run up to $1 million for a first violation and $3 million after that. Notably, both Anthropic and OpenAI supported the legislation.
The same day, Carnegie Mellon’s Software Engineering Institute announced its role in FLARE-AI, an open-source platform for reporting AI flaws and coordinating disclosure across developers, vendors, and government agencies. The premise is one insurance readers should sit with: a flaw discovered in one AI model may be quietly replicated across dozens of products built on the same underlying technology, and until now there has been no formal pathway to alert everyone affected. FLARE-AI routes standardized flaw reports into the same coordinated vulnerability disclosure machinery, including CVE identifiers, that the software security world has used for decades. Reports can flow to CERT/CC, the SEI’s AI Security Incident Response Team, and government clearinghouses.
Why This Matters for Insurance:
Carriers are not the audited entities under the Illinois law. Their foundation-model vendors are, and that changes the due diligence conversation. Starting in 2028, every major AI developer selling into your stack will be producing audit artifacts, incident reports, and catastrophic-risk disclosures on a statutory clock. Procurement teams and boards should be planning now to demand those documents in vendor negotiations, because a vendor’s 72-hour incident report to Illinois is material information for every carrier running that vendor’s model in claims or underwriting. The FLARE-AI premise sharpens the point: when carriers buy from a shared vendor stack, they share the vendor’s flaws, simultaneously and invisibly. One vulnerability in a widely deployed model is an accumulation event, not an isolated IT problem. The industry has spent two decades learning to think about correlated cyber exposure. Correlated model exposure is next, and the reporting infrastructure arriving now will determine how quickly anyone finds out.
Sources:
- Illinois Sets a New Standard for AI Oversight — Governing (Chicago Tribune/TNS)
- CMU Researchers Help Close a Critical Security Gap Across AI Platforms — Carnegie Mellon University
- FLARE-AI Platform
From the AI World: OpenAI Offers Washington 5% of Itself. The Vendor Dependency Question Just Grew a New Dimension.
The Financial Times reported on July 2 that OpenAI has proposed giving the U.S. government a 5% equity stake, worth roughly $42.6 billion at the company’s $852 billion March valuation. According to the reporting, Sam Altman pitched it as part of a broader arrangement in which Washington would hold 5% of each leading U.S. AI developer through a sovereign wealth fund vehicle, with the returns eventually distributed to the public. Whether Anthropic, Google, or Meta would participate is unclear; a source told CNBC that the administration and Anthropic have not discussed a stake.
The context explains the timing. Political pressure on the major labs has been building all year: public anxiety over AI-driven job losses, congressional interest in who captures AI’s economic upside, and national security scrutiny that saw the government order Anthropic’s most advanced models temporarily pulled offline in June over export control concerns. There is precedent for the mechanism, too. The administration already holds a stake in Intel and takes a revenue share on certain AI chip sales to China. Senator Bernie Sanders has proposed a more aggressive alternative: a one-time 50% tax on the shares of systemically important AI companies, deposited into a public fund. Any version of the OpenAI arrangement would likely require congressional approval, and nothing here is settled.
Why This Matters for Insurance:
Set aside the politics and consider the structural question for any carrier building on frontier AI. Every carrier AI deployment already involves a third-party infrastructure relationship the board may not fully understand. This proposal would add another party to that chain: the federal government as a shareholder in the vendor whose model runs your claims triage. That creates a genuinely novel governance puzzle. The same government that regulates how insurers use AI, through state insurance departments operating under frameworks shaped by federal pressure, would hold a financial interest in the companies supplying it. Whether that softens oversight, hardens it, or simply politicizes vendor selection is unknowable today. What is knowable: the ownership and control structure of the AI supply chain is now an active policy question, not a settled commercial fact, and carriers making decade-long platform bets should be pricing that instability into their vendor strategy.
Sources:
- OpenAI Proposes U.S. Government Own 5% Stake to Address Political Blowback — CNBC
- OpenAI Proposed Donating 5% of Its Equity to a US Sovereign Wealth Fund — TechCrunch
If You Found This Useful
This newsletter exists because insurance executives and independent agents need sourced, direct analysis on AI, not vendor marketing dressed up as news. If a colleague would benefit from receiving it, forward this issue or share the link.
And if you have not yet downloaded The Independent Insurance Agent’s AI Playbook, it is available free at insuranceindustry.ai/ai-playbook. More than 33,000 words of practical guidance, no cost, no catch.
Sources
- Ranking: Who Are the Insurance Industry’s AI Talent, Maturity Leaders? — Carrier Management
- Evident AI Index for Insurance — Evident Insights
- Bank of England Reviews AI Rules for Agentic AI in Finance — AI News
- Illinois Sets a New Standard for AI Oversight — Governing (Chicago Tribune/TNS)
- CMU Researchers Help Close a Critical Security Gap Across AI Platforms — Carnegie Mellon University
- FLARE-AI Platform
- OpenAI Proposes U.S. Government Own 5% Stake to Address Political Blowback — CNBC
- OpenAI Proposed Donating 5% of Its Equity to a US Sovereign Wealth Fund — TechCrunch
AI Disclaimer: This content was created with assistance from artificial intelligence technology. While content is based on factual information from the source material, readers should verify all details directly with the respective sources before making business decisions.
