The Deepfake Watchlist is Resemble AI's weekly surveillance of synthetic media incidents, ongoing cases, and disputed content shaping the news cycle. Each week we track confirmed incidents, emerging attack vectors, and claims under investigation, alongside the provenance, detection, and policy threads running underneath them. New to deepfakes? Start with our Deepfake 101 Guide.
★ Featured: Federal prosecutors say a three-person ring used AI-cloned CEOs and impersonated real lawyers to take more than $129 million
Federal prosecutors in the Southern District of New York unsealed an indictment on November 5, charging Jacob Soall, Sharon Hazan, and Guy Sheetrit with running a scheme that allegedly targeted thousands of businesses between July 2024 and February 2026, and the details kept surfacing through this week. The AP reported that victims included hospital systems in Missouri and California, a UK fintech, a Massachusetts mutual fund company, and a Florida pet food maker, and American Banker reported on November 9 that a Rhode Island bank lost about $2.68 million to the group in November 2024. All three defendants were arrested overseas and extradition requests are pending; Sheetrit's attorneys deny his involvement.
- Category: Fraud / Impersonation
- Type: Attack
- Modality: Audio, Video
- Policy / Regulatory: Each defendant faces seven counts, including wire fraud, aggravated identity theft, and money laundering; FinCEN has had a deepfake fraud alert for financial institutions on the books since November 2024.
- Trend: Executive impersonation fraud adding a second synthetic or stolen identity, a fake outside lawyer, to confirm the first.
- Attack vector: CEO spoofed by email, then a call with a "deal attorney" using a real lawyer's name, then a voice or video call with an AI-generated CEO authorizing the wire.
- What the indictment describes: Prosecutors laid out the playbook in steps that finance teams will recognize, for example:
- An email appearing to come from the CEO, asking an employee to sign an NDA for a confidential transaction
- A handoff to an "outside counsel" using the identity of a real attorney at the European office of a major international law firm
- A phone or video call in which AI mimicked the CEO's voice and sometimes the CEO's face
- Wire instructions routing the money to accounts the group controlled
Most of the advice companies give employees about CEO fraud is some version of "verify through a second channel," and this group built the second channel for them. The fake attorney carried the name of a real person at a real firm, someone an employee could look up and find, and by the time the cloned CEO got on the call there were already two voices telling the same story.
American Banker reports that a New York investment bank's back office stopped a $2.85 million transfer in October 2024, and that it came from a process control. The press release describes the defendants' tools as "super intelligence" technology, but nothing in the indictment points to anything more exotic than the voice and face cloning tools sold to consumers today, and that's the part that should worry any company with a public CEO, a public roster of outside counsel, and a finance team that has been trained to move fast when leadership asks.
1. Russian units allegedly used AI-generated videos to report captures of Ukrainian towns they never took
The Financial Times reported on November 8 that Russian military units and pro-war bloggers have been producing AI-generated or manipulated videos of soldiers raising flags in Ukrainian towns, and that inflated reports built on them travel up the chain of command to Vladimir Putin. Per The Daily Beast's account of the FT investigation, commanders told Putin they had captured Dobropillia in Donetsk Oblast and the Defense Ministry released drone footage of flag-waving soldiers, while Ukrainian forces still hold the town. The Institute for the Study of War has counted at least 35 suspect videos since the summer.
- Category: Political / Electoral
- Type: Attack
- Modality: Video
- Policy / Regulatory: No mechanism exists for verifying battlefield footage at the point of publication; analysts at ISW and the Conflict Intelligence Team are doing it after the fact.
- Trend: Synthetic media used to deceive one's own chain of command, not only a foreign audience.
- Attack vector: AI-generated or altered drone footage of flag-raisings, posted by unit channels and pro-war bloggers, with artifacts including a flag that reverses direction mid-clip and distorted bodies, per Novaya Gazeta Europe.
The tells described in the reporting were visible to open-source analysts without special tools. The problem is that the people receiving the footage had every reason to want it to be real.
2. Elvis's estate and Eminem's publisher sued Suno, putting Tennessee's ELVIS Act in front of a federal court
Rights holders for Elvis Presley, including Authentic Brands Group and Raleigh Music Publishing, and Eminem's publisher Eight Mile Style filed separate federal suits in California against AI music generator Suno, with the Eight Mile complaint filed November 6, Engadget reported, citing Variety. The Presley complaint invokes Tennessee's ELVIS Act and alleges Suno can produce tracks mimicking Presley's voice, while Eight Mile alleges copyright infringement across 269 compositions. Suno told Variety it does not allow unauthorized use of an artist's work and screens for misuse.
- Category: Brand / Likeness
- Type: Response
- Modality: Audio
- Policy / Regulatory: Tennessee's 2024 ELVIS Act protects an artist's voice as part of their likeness; this is one of its most prominent tests to date, alongside ongoing suits from Universal Music Group and Sony Music.
- Trend: Estates and publishers moving from takedown requests to litigation that targets the generator rather than individual uploads.
- Attack vector: Generative music models allegedly trained on copyrighted recordings and able to output vocals that imitate a specific artist.
Copyright protects the songs, while the ELVIS Act protects the sound of the person singing them, which is a much closer fit to what a voice clone actually takes, and a ruling on how far it reaches could set the tone for every other state weighing similar laws.
Warner settled its own fight with Suno last year through a licensing deal and whether a voice is something you license or something you can stop is the question these cases are going to force.
3. A former youth gymnastics and dance instructor was federally charged with using AI to make sexual imagery of children he photographed
Federal prosecutors in Rhode Island charged a 38-year-old North Smithfield man who worked at gymnastics, dance, and cheer facilities in Rhode Island and southeastern Massachusetts, WJAR reported on November 10. He allegedly photographed children at those facilities and used consumer AI tools to turn the images into sexual content. Charges include attempted production and receipt of child sexual abuse material and production of obscene visual representations of the sexual abuse of children. Homeland Security Investigations is contacting families.
- Category: CSAM / NCII
- Type: Response
- Modality: Image, Video
- Policy / Regulatory: Federal prosecutors are charging AI-generated abuse imagery under existing statutes, including a federal obscenity statute that reaches generated and drawn depictions of child sexual abuse.
- Trend: Ordinary photos of real children, taken in trusted settings, becoming the raw material for synthetic abuse imagery.
- Attack vector: Photos taken in person at youth activity facilities, processed through widely available consumer image and video generation tools.
What stands out from a policy view is how ordinary the inputs were: photos from a gym, taken by someone the families had reason to trust, run through tools anyone can download.
The charges show prosecutors are not waiting for AI-specific statutes to act. They also point to a hard problem for youth organizations, which now have to think about photography policies as a safeguarding issue in a way most of them never had to before.
4. Swatters are using AI voices to fool 911 dispatchers, and dispatchers say the tells are disappearing
NPR reported on November 11 that AI-generated voices are now part of swatting, citing an August bomb threat against Wheatland Union High School in Yuba County, California, that was made with a synthetic voice. Police responded overnight, classes were canceled, and no bomb was found. In a June survey of 766 public safety communications professionals by APCO International and the Public Safety Threat Alliance, about 14% said they had been targeted using AI tools.
- Category: Harassment / Public Safety
- Type: Attack
- Modality: Audio
- Policy / Regulatory: No law or regulation specific to AI-voice swatting; APCO is developing caller-questioning criteria, and dispatchers will keep treating every life-threatening report as real.
- Trend: Synthetic voices removing the audible cues dispatchers once used to spot hoax calls.
- Attack vector: AI-generated voice and background audio placed through police department or courthouse lines and transferred into 911 centers, which strips the location and caller data 911 would normally capture.
- What we heard in the reporting: APCO's Mel Maier said these voices used to be easy to recognize, and now insert natural pauses and filler words like "uh" and "let me think about it."
A forensic psychology professor quoted in the piece said he cloned his own voice with a commercial tool for under $20. The dispatchers' associations declined to describe their detection training in detail so as not to teach attackers, which is reasonable, and APCO's one public idea is to ask questions so granular that an AI can't answer them from public records or social media.
Dispatchers have to send help to anything that sounds like a life-or-death call, even when they suspect a hoax, so the attacker doesn't need to be convincing for long. A few seconds of believable panic is enough to send armed officers to a school.
Honorable mentions
AI images of missing pets used to extort reward money. A Somerville, Massachusetts woman whose cat went missing on Halloween received an AI-generated photo from someone claiming to have found it and asking about payment, Boston.com reported on November 6. The cat was later found safe.
Trump shared an AI cartoon of Dana White as press secretary, with a fighter body-slamming a reporter. The animated post went up on Truth Social over the weekend of November 8, and AP and Al Jazeera reported that Trump later said the job offer was a joke.
United Airlines banned third-party AI agents from booking flights. United said on November 11 it will restrict agent access across its digital platforms, citing credential security and federal rules that require in-person steps an AI check-in could skip.
The pattern
- Two of this week's stories are about the same weak point, a human being asked to decide in seconds whether a voice is real while the system around them rewards acting fast. The finance employee on a call with a cloned CEO and the 911 dispatcher hearing a synthetic panic call are in very similar positions, and in both cases the protection that actually worked, or is being built, is procedural: a back office that held a wire, a dispatcher trained to ask questions an AI can't answer.
- Synthetic media is increasingly aimed up a hierarchy rather than out at the public. The $129 million ring used a fake CEO to give orders downward, and the Russian battlefield videos allegedly carried fake good news upward to Putin. In both directions, the institution's own reporting lines did the work of making the content credible, because people tend to believe what arrives through official channels.
- The legal responses this week are converging on likeness as the thing that's harmed. The ELVIS Act suit treats Presley's voice as property his estate controls, and the Rhode Island charges rest on real children's photos turned into abuse imagery. Courts are being asked whether taking someone's face or voice is the injury in itself, regardless of what the output is used for, and the answer could shape the next round of state laws.
Watching next week
- Extraditions in the $129M case. Whether Spain and the UK move on extradition, and whether any of the victim institutions are named publicly as the case proceeds.
- Suno's response. How Suno answers the ELVIS Act claim specifically, since a motion to dismiss on that count would be the first real signal of how far the statute reaches.
- EU AI Act marking deadline. Providers have until early December to make AI-generated content machine-readable under the AI Act's transparency rules, RTÉ reported, and it's worth watching which platforms say they are ready.
- Rhode Island detention hearing. Whether the court holds the defendant pending trial, and whether more families come forward through HSI.
The Deepfake Watchlist publishes every Thursday. Subscribe to receive it in your inbox, or follow Zohaib Ahmed on LinkedIn for the weekly social companion. Track every documented incident in the Resemble Deepfake Incident Database, and read the full methodology in our 2026 Midyear Deepfake Threat Report.