You Can No Longer Trust What You See Online. Here’s What’s Being Done About It

Deepfakes have crossed from boardroom threat to everyday crisis — and the tools to fight back are finally catching up.

Not long ago, deepfakes were considered fringe cases – the kind of threat that warranted a line item in an enterprise risk report and a concerned memo from IT security. Executives worried about fraudulent wire transfers. PR teams drilled for impersonation scenarios. But for most people scrolling through their feeds, it felt distant. Abstract. Someone else’s problem. That era is over.

AI-generated images, voices, and videos have reached a scale and sophistication that makes them indistinguishable from reality for a majority of the population. According to Europol, as much as 90% of online content could contain some form of AI-generated material by the end of 2026. The question is no longer whether you’ll encounter synthetic media – it’s whether you’ll know it when you do.

1,567

verified deepfake incidents in 2025 alone

$1.3B+

in confirmed fraud losses tied to AI deepfakes

3.5 yrs

average news cycle for a corporate deepfake incident

Those numbers come from Resemble AI’s newly released Deepfake Threat Report, built from a proprietary database of 1,567 unique verified incidents drawn from 3,253 global news stories. And they almost certainly undercount the true damage: roughly 80% of incidents disclosed no financial loss figures at all.

The human cost behind the headlines

It’s tempting to reduce deepfakes to a fraud problem – wire transfers redirected, executives impersonated, stock prices manipulated. Those cases are real and serious. But the report’s data tells a more disturbing story about who bears the sharpest harm. Nonconsensual intimate imagery and child sexual abuse material accounted for 20% of verified incidents. These are not abstract corporate risk scenarios. They are crimes against real people, with consequences that outlast any news cycle.

For organizations, the reputational damage compounds over time. The average corporate deepfake incident remains in active news coverage for years – meaning a single synthetic video of a CEO can shadow a company’s brand long after it’s been debunked.

“We started by building voice AI models, so we understand how these systems work and how they can be weaponized. Multimodal generative AI security is now foundational for enterprises, employees and everyday people.”

— Zohaib Ahmed, CEO and Co-Founder, Resemble AI

Free tools for a public that needs them right now

Resemble AI, which has spent years on both sides of this problem – building generative voice models and developing detection infrastructure – has  released two free tools aimed squarely at the general public.

• Deepfake Detector for Google Chrome
  • One-click scanning of image, video, and audio across major platforms — X, TikTok, Instagram, Reddit, and more with color-coded results and frame-level video analysis.
• Deepfake Detection Bot for X
  • Tag @resemble_detect in any post with “is this fake?” and receive an automated authenticity verdict in-thread,  meeting misinformation where it spreads fastest.

The Chrome extension is particularly notable for its breadth: it handles audio segment-by-segment scoring and frame-by-frame video analysis, covering the range of synthetic media most people actually encounter.

Tools For Enterprise Organizations Under Pressure

For businesses operating in regulated industries or high-risk content environments, Resemble AI has also expanded its enterprise capabilities with three new features that address the specific frictions that have slowed adoption of deepfake detection at scale.

• Enterprise – Multimodal watermarking
  • Invisible signatures embedded at the moment of content creation, establishing tamper-resistant provenance across audio, video, and image at production volume.
• Enterprise – Zero retention mode
  • Submitted media is analyzed and immediately purged – no cloud storage, no data accessible to staff. Built for finance and healthcare environments where storage compliance is non-negotiable.
• Enterprise – Reverse image search
  • Cross-references submitted media against known debunked content and traces it back to original sources, surfacing propagation patterns for “zero-day” deepfakes with no prior digital footprint, even when provenance is deliberately obscured.

A Security Layer The Internet Never Had

What Resemble AI is building, incrementally, is something the internet was never designed to include: an authentication layer for media. Content on the web has always been easy to copy, edit, and redistribute without any record of what it originally was. Generative AI has simply made forgery so cheap and fast that the gap between creation and detection has become a crisis.

We’re entering a moment where seeing is no longer believing – and that puts the burden of judgment on every person and institution. The difference now is that verification tools are starting to meet people where misinformation actually spreads: in feeds, in threads, in real time. The question isn’t whether deepfakes will continue to proliferate. It’s whether we can make skepticism – and verification – as fast and frictionless as sharing.