What happens when a single AI-faked text “receipt” can land you in jail?

Melissa Sims, a Florida nurse, says she was arrested for allegedly violating a no-contact order after her ex-boyfriend used an AI-enabled fake text generator to create convincing “evidence.” Despite phone records later indicating no text was sent or received, prosecutors pursued the case for months—costing her nursing license, over $300,000, and months of her life. The no-contact charge was ultimately dropped, and she was later found not guilty on the original domestic violence charge.

Now she’s advocating for Melissa’s Law: legislation that would require metadata verification of digital evidence before arrests and create penalties for fabricating AI-generated/synthetic evidence.

We discuss:

* How AI-enabled text generators can create realistic “receipts”

* Why police and prosecutors often don’t authenticate digital evidence

* The technical detail that helped expose the fabrication (SMS vs iMessage)

* What Melissa’s Law would change in criminal evidence handling

* Why laws haven’t kept pace with synthetic media

* Why her ex has faced no consequences

Generative AI is not only about viral deepfakes. It also enables synthetic content at scale, including text, audio, and images that can be produced quickly and micro-targeted.

Ales Cap, a UCL PhD researcher studying AI and democracy, joins me to discuss what the evidence suggests about deepfakes in elections so far, why measuring impact is difficult without platform data, and why the biggest risks may be gradual through erosion of shared facts and trust. We also explore ideas for safeguarding electoral integrity and the promise of AI-enabled mass deliberation.

We cover:

* Deepfakes and election integrity

* Synthetic media beyond video

* The “infinite content machine” and micro-targeting

* Safeguarding elections and mass deliberation

Deepfake “nudify” apps are exploding online, but who’s building them, and how do they operate in plain sight?

Kolina Koltai, an open-source intelligence investigator at Bellingcat, traces the infrastructure that keeps these services profitable: shell companies, copycat domains, redirect sites, and payment workarounds, and how investigations can pressure operators into shutting down.

We discuss:

* The digital paper trail nudify operators leave behind

* How these services evade app stores, payment processors, and hosting platforms

* Why targeting users misses the bigger problem: site owners and profit incentives

* The idea of “friction,” and why making abuse harder matters even if you can’t stop it completely

* The cultural shift needed to prevent this from becoming normalized

In 1969, President Nixon had a speech ready in case Apollo 11 ended in tragedy. Fifty years later, artists Francesca Panetta and Halsey Burgund brought that speech to life using deepfake video and early AI voice cloning, showing Nixon delivering words he never actually spoke.

Their Emmy Award-winning interactive documentary In Event of Moon Disaster is both a haunting alternate history and a media literacy experiment about how easily the past can be rewritten. We go behind the scenes on the tech, the ethics, and what responsible synthetic media can look like.

We discuss:

* How the project came together at Harvard and MIT

* Why Apollo 11 was the perfect vehicle for this message

* The ethics of demonstrating deepfake danger without becoming misinformation

* Where deepfake detection and media literacy are headed, and their limits

See the project: moondisaster.org

Deepfakes are supposed to destroy our ability to trust video evidence—but what if that fear rests on flawed assumptions about how we gain knowledge in the first place?

Philosopher Dr. Joshua Habgood-Coote (University of Leeds) argues that video has never been a simple window into reality. It’s always been interpersonal knowledge, dependent on trusting the people and institutions behind what was filmed, edited, and shared. In his 2023 paper “Deepfakes and the Epistemic Apocalypse,” he challenges the panic and draws surprising parallels to Victorian spirit photography.

We discuss:

* Why “seeing is believing” was never quite true for video

* How 19th-century newspapers handled widespread photo faking through social norms, not tech

* Why the real harms of deepfakes (including non-consensual intimate imagery) are about power and harassment, not just deception

* What a society that’s adapted to deepfakes might actually look like

Content note: This episode discusses non-consensual intimate imagery and sexual abuse.

Noelle Martin was 18 when she discovered doctored intimate images of herself across dozens of porn sites. When she spoke out, perpetrators escalated to deepfake videos. A decade later, she’s an Australian lawyer fighting for legal reform—while still being targeted.

We discuss:

* Why takedowns are often futile, and how perpetrators escalate

* The gap between criminalizing distribution vs. creation

* What Australia’s 2024 reforms do and don’t solve

* Why this requires a coordinated global response

* Noelle’s advice for others experiencing NCII abuse

Smartphone footage has transformed how we document war crimes. But what happens when anyone can claim “it’s a deepfake”?

Professor Yvonne McDermott-Rees (Swansea University) leads the TRUE Project, investigating how AI-generated media affects trust in user-generated evidence of human rights violations.

In this episode, we explore:

• Whether the “liar’s dividend” is actually showing up in courtrooms

• What mock jury experiments reveal about how people evaluate digital evidence

• Why the biggest risk may be real footage being dismissed—not fakes entering court

Learn more: trueproject.co.uk