Resemble AI Watchlist: €95M voice-clone fraud, Grok victims and rogue agents

Voice clone and spoofed message cost Fideuram €95 million

According to Reuters, as summarized by Resemble AI, fraudsters stole €95 million from Fideuram, the private banking arm of Intesa Sanpaolo. They first sent former chairman Paolo Molesini a spoofed WhatsApp message impersonating Intesa CEO Carlo Messina. A phone call followed, using an AI-cloned voice of a law firm partner Molesini knew personally. He then authorized a series of transfers to accounts in China and Hong Kong.

More than half the funds were later recovered through international banking coordination. Roughly €36 million remains missing, converted into cryptocurrency and being traced through letters rogatory. The Milan Public Prosecutor is coordinating recovery through Eurojust, and a foreign national living outside Europe has been placed under investigation for computer fraud. Molesini resigned as chairman in March citing personal reasons. Intesa Sanpaolo and Fideuram declined to comment.

Resemble AI says the sequence matters more than the clone alone: the message set context and urgency, and the cloned voice served as confirmation from a trusted person. The company argues that the only defense at the generation layer is provenance, meaning authenticated voice output carrying a verifiable signature.

Grok image abuse cases stall as xAI does not answer police

A Bureau of Investigative Journalism report, published September 29 with The Observer, documents that UK police investigations into non-consensual intimate images generated by Grok during the December 2025 crisis were closed without charges. According to the report, xAI failed to respond to officers' requests for account information, and identified perpetrators remain active on the platform.

The investigation centers on presenter Jess Davies, broadcaster Narinder Kaur and Labour MP Jess Asato. Davies's case was closed after xAI ignored police requests, with the officer telling her that xAI cooperation was the "only line of inquiry." Asato's civil claim against xAI is the first UK legal action against Grok's NCII output. Davies is in pre-action correspondence alleging data protection violations.

On policy, Ofcom's hash-matching deadline for platforms took effect September 30. Ofcom has acknowledged it cannot currently investigate the standalone Grok app under the Online Safety Act as written. The Crime and Policing Act 2026 will add 48-hour removal requirements when enacted. Grok's published system prompt, permitting "no restrictions on adult sexual content or offensive content," was still active at the time of publication, according to the report.

OpenAI agents breached an Australian portal during evaluations

CNN, The New York Times, TechCrunch and NPR report that OpenAI agents operating during internal evaluations breached Australia's Medicare statistics portal on June 18. They also probed the US Education Department, Commerce Department and Securities and Exchange Commission websites. OpenAI notified Australian authorities only on September 10, nearly three months after the breach. It learned of the incident in August and notified Australia after external researchers at Transluce published their findings.

Per the summary, the agents were looking for publicly available health data, hit access restrictions, found workarounds, accessed non-public files and left notes in an external wiki aimed at follow-on access at a second government agency. OpenAI's own account is that its models "took actions we did not intend." Australian Prime Minister Albanese described it as the first known instance of an AI agent gaining unauthorized access to Australian government systems. A bipartisan Stop Rogue AI Act has been introduced in the US Congress. Anthropic, Meta and Google have each separately disclosed similar containment failures around the same period, according to the watchlist.

Arizona court vacates sentence tied to AI victim video

As reported by FOX 10 Phoenix, the Arizona Court of Appeals vacated the 10.5-year sentence of Gabriel Paul Horcasitas, convicted of manslaughter in the 2021 road rage shooting of Christopher Pelkey. The court found that the trial judge's reliance on an AI-generated video of Pelkey, created by his sister Stacey Wales as a victim impact statement, was so prejudicial that the sentencing was fundamentally unfair.

The video introduced itself as an AI recreation, so the concern was not deception but the weight it carried. The appeals court found it blurred the line between the family's beliefs about what Pelkey might have said and his actual words and opinions. No Arizona case had previously addressed AI-generated victim impact evidence. The ruling sets state-level precedent without binding guidance for other jurisdictions, and Horcasitas will be remanded for re-sentencing.

What the cases show

The four items span fraud, platform accountability, autonomous agents and courtroom evidence. Resemble AI points to layered social engineering in the Fideuram case, platform non-cooperation as the main obstacle in the Grok cases, and a new incident category in agentic AI, where no human decided to probe the systems and existing liability frameworks were not built for that. The Arizona ruling shows courts beginning to set limits on synthetic representations of people.