Insurers face a high-stakes AI arms race as risks and opportunities collide
Artificial intelligence is forcing a reset in how carriers and brokers think about exposure. James Bullock-Webster, director and head of tech, media and cyber at New Dawn Risk, said the pace of adoption has been "incredibly powerful" - and those who fail to adapt risk being left behind.
"We're all living it," he said. "There are known unknowns and unknown unknowns with AI. Insurers know they have to get arms around it, be prepared to have that risk transferred to them in order to remain relevant."
The tension is clear: AI can both strengthen defenses and enable attacks. That creates an arms race between threat actors and legitimate businesses - one insurers must track and price in real time.
The risk picture: exposures you can price today
- Algorithmic bias and discrimination: Underwriting and claims automation can misfire, producing unfair outcomes that trigger complaints, regulatory scrutiny, or class actions.
- Data privacy and model leakage: Training on sensitive data or poor retention controls can lead to privacy breaches and regulatory penalties.
- AI-boosted cyber threats: Faster phishing, deepfake voice/video, automated vuln discovery, and code-gen for malware increase frequency and severity.
- Generative content risk: Deepfakes, misinformation, defamation, and copyright claims tied to model outputs and training data.
- Systemic/aggregation risk: Concentration in common models, libraries, or vendors can create correlated loss events across portfolios.
Market structure: speed, scale, and where MGAs fit
Smaller players like MGAs can move faster without legacy tech drag. Global carriers bring capital, breadth, and balance sheet strength to backstop emerging risks. Both are needed.
From the Lloyd's market view, MGAs act as accelerators. "Probably about 40% of all capacity and premium that's collected at Lloyd's is done through third parties like MGAs," Bullock-Webster said. Delegated authority gets specialty capacity into niche segments efficiently and extends reach.
Lloyd's syndicates often have more freedom on wordings than tightly constrained domestic carriers in North America. That autonomy, combined with MGA partnerships, can speed up product evolution for AI risks.
AI inside carriers: a double-edged sword
Using AI in underwriting and claims can cut costs - and create fresh liabilities. "If these systems make mistakes, like wrongly denying a claim or unintentionally discriminating, insurers could face investigations or lawsuits," he said.
The fix isn't hype. It's controls: clear governance, documented decision logic, audit trails, and human oversight at points of high impact. If you can't explain it, you can't defend it.
A practical playbook for carriers
- Map your AI footprint: Inventory models, use cases, and vendors across underwriting, claims, SIU, and distribution. Set ownership and KPIs.
- Adopt a control framework: Bias testing, validation, monitoring, and incident response aligned to recognized standards like the NIST AI Risk Management Framework.
- Tune product strategy: Cyber updates for deepfake/BEC loss, media liability for generative content, tech E&O for model providers, and endorsements addressing training data practices.
- Wording levers: Data provenance warranties, explainability and audit rights, human-in-the-loop requirements, model change notifications, and clear definitions of "AI services."
- Aggregation management: Track dependencies on major models and key vendors; run stress tests and secure aligned reinsurance support.
- Claims readiness: Protocols for verifying deepfakes, preserving model logs, expert panels for AI forensics, and fast-lane guidance for adjusters.
- Pricing signals: Credit insureds for governance, secure development, prompt logging, red-teaming, and vendor due diligence. Penalize "black box" deployments at scale.
- Regulatory posture: Document decision rationale, fairness reviews, and customer disclosures. Train staff - especially for adverse action scenarios.
Actions for brokers and MGAs
- Ask process questions, not just product questions: How does the carrier use AI in triage, pricing, and claims? What controls back it up?
- Clarify coverage boundaries: Explain where cyber ends and media/tech E&O begins for AI-driven loss. Set expectations before a dispute does it for you.
- Expect new E&O angles: Advisory missteps around AI use or coverage fit can come back as professional liability claims.
- Build niche solutions: Fast-track delegated products for AI-native risks (model ops, data labeling, synthetic media studios) with precise wordings.
- Upskill the team: The market will reward brokers who can read an AI control environment and translate it into price and terms. Consider structured learning paths; see AI courses by job.
The takeaway
"AI is here and the genie's out of the bottle," Bullock-Webster said. Insurers that build product, governance, and claims muscle now will win clients and relevance.
Wait too long, and others will set the terms - and take the business.
Your membership also unlocks:
