It turns out P&C insurers aren’t the only ones harnessing generative AI to modernize operations—so are the fraudsters out to defraud them.

Carriers in the UK, for example, recently reported a 300% increase in auto claims supported by AI-manipulated vehicle images over the past year. According to InsuranceTimes, claims teams are also seeing an increase in claims documents produced using generative AI—including fake foreign hospital bills used in travel insurance scams.

While “shallowfake” images created using tools like Photoshop have long been a problem, deepfakes are particularly troubling. This form of generative AI makes it easier than ever to produce synthetic yet highly realistic images, video, audio, and documents. Using widely available tools like DALL-E or Midjourney, for instance, scammers can file claims for fraudulent damage or open new policies for assets that don’t even exist.

Cybercriminal networks aren’t the only culprits, either. Sometimes, ordinary policyholders rationalize filing fraudulent claims as compensation for years of premium payments. But now, expanding access to generative AI tools is fueling an escalating arms race between insurers and fraudsters. And the stakes are growing fast.

Counting the Cost of Deepfake Attacks in P&C

It’s estimated that the worldwide impact of P&C insurance fraud in all its forms tops $120 billion annually, with 10% of all claims believed to be fraudulent. In the US, P&C (home, auto, commercial, etc.) accounts for roughly $45 billion of those annual losses, according to the Coalition Against Insurance Fraud (CAIF). Ultimately, these crimes cost families up to $700 in extra premiums yearly. But the US is hardly alone.

Determining the exact percentage of fraudulent claims involving deepfakes and other forms of manipulated content is hard to pin down. But a recent industry survey found that 19% of claims handlers believe AI and other digital tools factor into 25% of fraudulent claims cases, and 64% believe AI plays a role in 5% to 10% of all cases.

Where the Deepfake Threat Bites Hardest

• • Low-severity, high-volume lines: Auto glass, weather-related roof repairs, and minor water damage claims rely on photo-only adjudication, making them easy targets

• • Digital-first direct carriers: Self-service FNOL offers a competitive edge, but it also lowers the friction that might otherwise deter fraudsters

• • Catastrophe response surges: Post-event spikes can overwhelm adjusters; deepfake images can slip through while human attention is thin

The Expanding Deepfake Attack Surface

While deepfake images used in policy and claims fraud generate a lot of headlines, it’s not the only way fraudsters leverage this and other forms of generative AI to scam insurers.

According to Pindrop’s 2025 Voice Intelligence & Security Report, there was a 475% surge in synthetic-voice attacks targeting insurance contact centers last year. In these schemes, callers using cloned voices redirect claims payments or fool adjusters into revealing login credentials.

Meanwhile, IBM’s 2025 Data Breach Report estimates that deepfakes factor into 35% of all AI-enabled social engineering attacks aimed at infiltrating corporate networks. The average cost of a successful breach for US companies jumped 9% over the past year to an all-time high of $10.22 million per incident, while the global average hit $4.44 million. Swiss Re warns that deepfakes and other forms of malicious AI are “eroding trust and inflating operating costs” for carriers.

Four Strategic Plays to Gain the Upper Hand

Tipping the balance of power in such an asymmetric arms race won’t be easy. But here’s how P&C insurers can leverage these same and other AI technologies to fight back:

Re-platform to a cloud-native, API-first core system

Carriers operating with a modern, cloud-based insurance platform like Guidewire, for instance, can integrate solutions that detect modified or fabricated visuals and content from a wide array of ecosystem partners, including those from Truuth and Photocert. As Boston Consulting Group points out, an API-first infrastructure enables carriers to rapidly plug defenses against AI-wielding fraudsters directly into workflows.

Deploy multimodal AI defenses at every checkpoint

With a modern architecture, carriers can also harness solutions that utilize computer vision, audio forensics, and large-language anomaly detection across the insurance value chain—from policy submission processes to claims intake, triage, payment, and beyond. By some estimates, combining image, text, and voice telemetry (or multimodal analysis) can raise deepfake catch rates to 98% or more. According to Deloitte, the savings could top $160 billion by 2032.

Adopt real-time liveness testing and behavioral analytics

Live video capture and behavioral biometrics make it nearly impossible for false claimants to stay believable throughout the claim life cycle. With Guidewire, for instance, integrated solutions like those from Livegenic, Ravin, and U Scope enable carriers to utilize live photo, video, or audio to verify claims information and collaborate with policyholders. And the advanced analytics capabilities built into the platform interrogate high-resolution internal and external data sources to detect behavioral patterns and other factors that may signal fraud.

Fuse cybersecurity and fraud teams

In the age of deepfakes, disjointed fraud prevention efforts are giving way to integrated teams that view fraud, cyber, and analytics as a single theater—linking image forensics, network security, and claims metadata in a shared threat cockpit. While not limited to malicious AI, Forrester Research estimates that organizations with combined cybersecurity and fraud teams are 40% more likely to prevent fraud from becoming financial loss.

Seeing Is Deceiving–But Deepfakes Can Be Defeated

Generative AI in P&C insurance is here to stay. But so are the adversaries brandishing the same technology. Expect regulatory scrutiny to intensify, consortium and ecosystem data-sharing to accelerate, and deepfake defenses to grow more seamless. Carriers who invest in cloud-native core systems, multimodal detection, and empowered fraud and cybersecurity teams won’t just blunt the deepfake fraud wave. They’ll also unlock faster, more accurate settlements, and enhance the customer experience for honest policyholders.