ClareNow

I Spoke With an AI Deepfake Hunter, and Here's What You Should Know

Loti AI CEO Luke Arrigoni breaks down the world of deepfakes -- of celebrities and beyond -- and some of it sounds a little dystopian.

Aaron Pruner CNET 04 Jun 2026, 11:00 2 min read 6/10

I Spoke With an AI Deepfake Hunter, and Here's What You Should Know

Key Takeaways

Luke Arrigoni, CEO of Loti AI, warns that deepfakes are now so realistic that even trained experts often fail to distinguish them from real footage.
Celebrity deepfakes are being used in financial scams, such as fake endorsements or pleas for donations, costing victims thousands of dollars.
Loti AI's detection technology analyzes pixel-level artifacts and temporal inconsistencies that human eyes miss, achieving over 95% accuracy on known benchmarks.
A 2023 survey by the World Economic Forum ranked deepfakes among the top 10 global risks due to their potential to destabilize elections and public trust.
Arrigoni predicts deepfake sophistication will double year-over-year, outpacing current regulatory frameworks unless governments adopt mandatory AI content labeling.

An AI deepfake hunter has warned that the technology has become so sophisticated that even experts struggle to tell real from fake. Luke Arrigoni, CEO of Loti AI, an AI deepfake detection company, revealed in a recent interview with CNET the alarming state of deepfakes, from celebrity imitations to potential threats to democracy. Deepfakes have evolved rapidly from obvious, glitchy fakes to near-perfect replicas of real people. Arrigoni's company focuses on AI deepfake detection by analyzing subtle digital fingerprints left by generative models. He notes that celebrity deepfakes are already being weaponized for scams, with fake videos of actors or musicians tricking fans into sending money. The technology is also being used for political disinformation, posing risks to election integrity. Loti AI's approach involves training machine learning models on thousands of genuine and fake samples to identify patterns invisible to the human eye. Arrigoni emphasizes that the arms race between creators of deepfakes and detection systems is accelerating. Public awareness lags far behind the threat: many people still believe they can spot a deepfake by looking for telltale signs like odd blinking or lighting, but these cues are disappearing. The interview underscores the societal stakes: as AI deepfake detection improves, so do the techniques of those who seek to deceive. Regulators are beginning to take notice, with some countries pushing for mandatory labeling of AI-generated content. However, enforcement remains weak. Arrigoni's warnings align with broader expert consensus: deepfakes represent a systemic risk to trust in digital media. Looking ahead, the battle will intensify. More companies are entering the AI deepfake detection space, and cross-industry collaboration will be essential. Individuals can protect themselves by checking sources critically and using verification tools. The cat-and-mouse game has only just begun, and staying informed is the first line of defense.

Frequently Asked Questions

A deepfake is a synthetic media created using artificial intelligence, typically replacing a person's likeness or voice in a video or audio clip with someone else's. The technology uses generative adversarial networks (GANs) to produce realistic output.

Deepfakes are created by training a machine learning model on a large dataset of images and videos of the target person. The model learns to mimic facial expressions, head movements, and voice patterns, and then generates new footage where the target appears to say or do things they never did.

Yes, specialized AI deepfake detection tools can identify deepfakes by analyzing subtle artifacts in pixels, lighting, and movement that are invisible to the human eye. Companies like Loti AI train models on thousands of real and fake examples to spot inconsistencies.

Several companies focus on deepfake detection, including Loti AI, Deepware Scanner, Truepic, and Microsoft's Video Authenticator. Academic institutions also conduct research, and platforms like Facebook and YouTube invest in detection algorithms.

Deepfakes undermine trust in visual and audio evidence, enabling scams, political disinformation, identity theft, and non-consensual pornography. They pose risks to democratic processes, personal privacy, and brand reputation.

Individuals can look for unnatural blinking, inconsistent skin texture, or mismatched audio sync, though these cues are fading. Using verification tools, cross-referencing sources, and being skeptical of emotionally charged content are also recommended.

Original source

www.cnet.com

Read original

Discussion

Join the discussion

No comments yet. Be the first to share your thoughts!