AI-Doctored Claim Photos: Fraud Checks for Kenyan Insurers

Insurance fraud reports in Kenya are rising, and generative AI has made one of the oldest fraud tools, the doctored photo, much easier to produce. A claims team that treats a photo as proof is now exposed. The fix starts with the intake process, not new software.

What the Kenyan numbers show

According to Sharp Daily's reading of Insurance Regulatory Authority data, the IRA's fraud investigation unit received 52 cases in the first quarter of 2026, up from 35 a year earlier. Theft by insurance agents jumped from 2 cases to 17, the largest category that quarter.

Medical claims made up 46.2% of incurred claims in that quarter, with private motor at 21.4% and commercial motor at 19.1%. Earlier fraud unit data for the first half of 2025 showed 45 of 85 fraud reports involved motor insurance. Common patterns included fake accident claims, forged certificates, falsified theft reports and fabricated medical bills, as Eastleigh Voice reported.

What AI changes about a claim photo

Law firm Debevoise warned in January 2026 that generative AI tools can produce photorealistic images, or take a real photo of minor damage and exaggerate it. The result can be a payout for damage that never happened, or for far more than the real repair cost. Debevoise gives no figures on how often this happens, so treat it as a growing risk and not a measured one.

Three checks to add this week

Several photos with a reference object: Ask for photos from different angles and distances, with an object in frame that shows scale.
Original files: Ask for the original photo or document, not a screenshot or a re-saved copy.
A short live video call for larger claims: An adjuster asks the claimant to pan, zoom and show specific features in real time.

Do not tighten so far that honest claimants wait weeks. Debevoise cautions that overly strict controls can slow or wrongly deny legitimate claims.

Where AI and automation can help

Once intake is tighter, repeat checking is where automation earns its place. Sharp Daily notes that insurers can use AI, machine learning and data analytics to spot unusual patterns and flag suspect claims before payment, while stressing that technology alone cannot eliminate fraud. Three practical uses:

Claim triage: Score incoming claims for patterns such as repeat claimants, the same repairer or outlier medical bills, and route flagged files to a senior officer.
Image and file checks: Flag reused photos or images that look AI-generated, for a person to review.
Agent premium reconciliation: Match premiums agents collect against policy records and bank or M-Pesa lines, so gaps show up early. With agent theft the largest fraud category, this matters for brokers.

Keep a person in charge

Claim files hold personal and health data, which the Data Protection Act 2019 covers. Use business grade tools, upload only what a check needs, and log what the AI flagged. Never accuse a claimant on a detector score alone, because detection tools can be wrong. A person should make every denial or fraud referral.

What this means for insurers and brokers

Swiss Re reviewed 187 reported AI use cases at insurers and found fewer than 5% had disclosed a measurable financial impact. So treat AI as something to test on one process and measure, not a promise.

Start with one change. Ask for several photos with a scale object on your next motor claim, and track how many files need follow-up over a month. Then decide whether automation is worth adding. Brokers and agents can also share the checklist with clients, so honest claimants know what to send, which helps settlement move faster.