With artificial intelligence-generated image tools getting more accessible, users are creating realistic and sometimes flawless recreations, sometimes to even make fake claims, experts said. This could spell real financial loss for companies, including in the ecommerce and quick commerce space, they said.

Such an instance occurred in November, when a quick commerce platform user allegedly created an AI-generated image of a tray with cracked eggs to claim a refund for an order. This incident suggests companies across the board that depend on visual evidence as proof of incident or transaction could suffer losses.

“Only the platforms that upgrade their trust and safety stack will survive the fraud arms race,” said Ami Kumar, cofounder of deepfake and manipulation detection system Contrails AI. “Most support workflows today have no deepfake detection, manipulation forensics, saliency checks or a way to validate the authenticity of customer- or seller-submitted media. This becomes particularly important in the case of quick commerce or ecommerce where AI-amplified refund fraud could destroy the unit economics.”

Companies, regardless of the sector, are waking up to the risks that come with such fabricated claims. Even the routine filing of reimbursement claims at a company or submission of insurance claims could be further scrutinised in 2026. Some said the age of attestation and stamping may make a comeback.

“We have started demanding that physical bills be submitted along with the digital bills,” a senior company executive at a manufacturing firm who did not wish to be named said. “We had stopped asking for physical bills but recently we had a few incidents where employees were submitting fabricated bills and we’ve enforced this now as a policy.”

The problem extends well beyond retail, said Rahul Sasi, cofounder of cybersecurity firm CloudSEK which has built a free deepfake analyser. He believed that insurance, logistics, banking, compliance, construction and even internal expense management sectors will face similar risks as visual ‘proof’ becomes cheap to fabricate.

“Companies won’t fully go back to paper, but they will add verification layers like metadata checks, provenance and selective human review to restore trust in digital evidence,” Sasi said. “Technically, platforms need automated manipulation detection, risk-based escalation, and provenance-aware capture and upload flows. In a world where ‘seeing is believing’ no longer holds; verification becomes core infrastructure, not a nice-to-have.”

With AI models learning rapidly, the erosion of trust when it comes to photographic evidence has been immense and some felt this would only lead to adoption of old measures to verify again.

“AI has turned the notion of trust attached to image-based evidence on its head,” said Dhruv Garg, founding director of Indian Governance and Policy Project. “It brings in serious, real-world challenges for companies but also for more critical functions too. Something like the e-KYC too may be manipulated using AI tools in the near future and that is a real concern,” he said. “We might enter a time where organisations will have to adopt old-school methods based on human in the loop to verify authenticity like the way banks used to earlier send agents to a customer’s home or office.”

However, Garg too felt that provenance-based authentication and blockchain technology may hold some solutions.

“Within the justice system, there are still checks and balances available,” Garg said. “(But) in general, when it comes to the concept of photographic proof in daily use cases, the credibility attached to it may face challenges in the short term.”

He added that many of the processes that digitisation was meant to solve might need further human verification, at least in the interim.

“There are also coalitions like the C2PA (Coalition for Content Provenance and Authenticity) which are actively working with AI companies to bring in a system to label content that is AI-generated with a small signature,” Kumar said. “Measures like this could help too but in the short term, especially in the Indian context, it is plausible that companies will put in place mechanisms for human verification.”

But in the long run, he reiterated that it will only be computer vision, deepfake detection tools and other technology that could provide a solution to this issue at scale.

Join the community of 2M+ industry professionals.

Subscribe to Newsletter to get latest insights & analysis in your inbox.

All about ETCISO industry right on your smartphone!