Sept 2026Event location unknownUnidentified image tool
Der Standard reported on 28 September 2026, from an interview with Lena Schilling (Green MEP, Austria) conducted with the Salzburger Nachrichten, that about two weeks earlier she had been informed, through an ongoing survey by the Berlin think tank Agora Digitale Transformation, that websites with AI-generated nude images of her existed; about a dozen websites showed such deepfakes or linked to them. She says photos of her were used on AI platforms to undress her against her will, that she felt sick on first seeing the images and then angry, and that she and her team documented the cases, filed a complaint with the Austrian police, reported them to the anti-hate body ZARA, wrote to the websites and reported them to search engines; all sites were offline after about ten days. Through a Greens statement relayed by APA (Krone, Salzburger Nachrichten) she said 'Man hat uns gegen unseren Willen ausgezogen' and called for political decisions to protect women. Her case belongs to a second, still ongoing wave of the Agora study, whose first wave found sexualised deepfakes of 138 female and nine male national parliamentarians in 22 EU countries.
Core concern Medium reported severity Media Coverage
AI involvement reported · Causal attribution supported · 4 sources, 2 underlying accounts · Added 30/09/2026
23 Feb 2026 to 25 Feb 2026ItalyUnidentified voice-cloning tool
On 23 February 2026 Paolo Molesini, then chairman of Fideuram (the private-banking subsidiary of Intesa Sanpaolo), received a WhatsApp message from an unknown number from someone claiming to be Intesa's chief executive Carlo Messina, announcing a confidential acquisition that had to be executed through Fideuram. The same day a caller presenting as a lawyer of A&O Shearman whom Molesini knew, whose voice ANSA, Il Fatto Quotidiano and Corriere della Sera, reporting from the Milan court papers, describe as created with artificial intelligence (today.it, which also cites the seizure decree, writes that the court papers do not yet answer whether the voice was imitated), had him sign a confidentiality agreement and sent eleven payment instructions; Fideuram's treasury head was separately contacted by someone posing as Fideuram's CEO. Between 23 and 25 February eleven transfers totalling about 95 million euros went to accounts in Portugal and at Bank of China; the bank's alarm systems and the Milan prosecutors recovered about 40-42 million from China and 13 million seized in Portugal, leaving at least 36 million (39.5 million per the court papers) missing after conversion into cryptocurrency. Molesini, who Corriere writes is not under investigation, resigned as chairman on 12 March 2026, which Fideuram announced as being for personal reasons; a 48-year-old Israeli citizen is under investigation as a member of the gang.
Core concern Medium reported severity Investigation Opened
AI involvement reported · Causal attribution supported · 8 sources, 5 underlying accounts · Added 30/09/2026
17 Apr 2026 to 31 May 2026Event location unknownMeta AI support assistant
Meta announced the rollout of its AI support assistant on Facebook and Instagram on 19 March 2026. Meta's filing with the Maine Attorney General (notice dated 5 June 2026) says unauthorized third parties exploited a vulnerability in its AI-assisted Instagram account recovery tool (High Touch Support) to receive password reset links for accounts they did not own, and the listing gives 17 April 2026 as the breach date and 31 May 2026 as the discovery date. Videos that attackers posted, as described by TechCrunch, 404 Media and Krebs on Security, show attackers asking the assistant in a chat to link a new email address to a target username. Reported victims include the security researcher Jane Manchun Wong, who posted that her password was changed without her knowledge, the Instagram account of the U.S. Space Force's Chief Master Sergeant, and the accounts of Sephora and a dormant Obama White House page (TechCrunch marks the last as disputed by Meta). Krebs and the BBC report pro-Iran defacement of some accounts, and TechCrunch reports that some victims were locked out and that short handles were offered for resale. The filing gives 20225 persons affected in total and 30 in Maine, and the notice calls the Maine figure an upper bound. Meta says the tool worked as intended, that a bug in a separate code path failed to check the email address, and (through a spokesperson to Gizmodo) that the failure was not due to the AI agent itself. Meta says it disabled the tool on 31 May 2026, the day it discovered the exploitation.
Contextual tracker case Low reported severity
AI involvement supported · Causal attribution disputed · 12 sources, 2 underlying accounts · Added 29/09/2026
30 Mar 2026South AfricaSASSA eLife facial verification
The South African Social Security Agency (SASSA) introduced an online eLife Certification (life certification) for grant beneficiaries that uses biometric verification through its electronic Know Your Client (eKYC) system. IOL reports the certification was implemented on 30 March 2026. SASSA says beneficiaries who do not complete life certification as directed may face payment delays or suspension. On 10 April 2026 SASSA apologised to beneficiaries who could not access the portal, said system glitches linked to interfaces with other departments had caused delays, disruptions and long queues at its offices, and said the problem was resolved. On 23 April 2026 IOL reported that a pensioner couple said they had tried the facial recognition option 22 times since 2 April without success, and that beneficiaries nationwide told IOL they could not complete the certification, citing failures with facial recognition and one-time PINs, with one pensioner also reporting a message that Home Affairs was not available to verify their particulars. A SASSA spokesperson said the portals work and that 13,644 (88%) of the 15,499 unique clients who accessed the online verification services by 16 April were verified, and IOL reports SASSA admitted the system has been working intermittently. In a May 2026 report on a parliamentary reply, IOL said SASSA stated that unsuccessful facial recognition attempts on online platforms were among the causes of non-verification (those beneficiaries are redirected to fingerprint checks at local offices) and that it had recorded 7,779 complaints linked to its electronic facial biometric system. The department attributed facial verification issues to poor lighting, unstable connectivity or missing biometric records at Home Affairs. Neither May report mentions the eLife portal, and IOL places the figures within a biometric verification rollout that it dates from September 2025. The reports do not say how many grants were suspended because of facial verification failures.
Contextual tracker case Low reported severity
AI involvement reported · Causal attribution alleged · 5 sources, 4 underlying accounts · Added 29/09/2026
10 Jul 2026Event location unknownCodex coding agent
On 10 July 2026 Matt Shumer, founder and chief executive of the AI start-up OthersideAI, posted on X that OpenAI's newly released GPT-5.6 Sol "just accidentally deleted almost ALL of my Mac’s files". Accounts of his post-mortem say he was testing a high-autonomy multi-agent "Ultra mode" at OpenAI's invitation, with the Codex agent given Full Access to his machine; during a cleanup task a sub-agent expanded $HOME incorrectly and ran a recursive delete of his home directory, he noticed a problem about 81 minutes into the session, and by the time he stopped the process most of its contents were gone. Three days later he wrote that the deletion "absolutely sucked" and that OpenAI staff, including Greg Brockman, had contacted him to help. OpenAI confirmed, as reported by The Register, that GPT-5.6 had deleted users' files without authorisation, describing it as an "honest mistake" that usually occurred in Full-Access mode without sandboxing, and said it was adding safeguards. Whether the files were recovered is not reported.
Core concern Low reported severity Media Coverage
AI involvement supported · Causal attribution supported · 6 sources, 2 underlying accounts · Added 29/09/2026
19 Aug 2026FranceUnidentified AI tool
According to Gossement Avocats, which quotes the order, and the court's vice-president on France Inter, an order of 19 August 2026 (n°2508168) of the tribunal administratif de Rennes rejected a claimant’s application contesting a VAT reassessment and fined the claimant 500 euros for an abusive application. As quoted by Gossement Avocats, the order says the application had manifestly been written with an artificial intelligence tool and that its grounds lacked the most elementary details needed to assess them. The court also found it inadmissible because no prior complaint to the tax service was shown, and noted that it repeated a request with substantially the same object rejected in October 2025. The court’s vice-president cited the case on France Inter; commentators stress that AI use alone was not the basis for the fine.
Core concern Low reported severity Regulatory Action
AI involvement reported · Causal attribution supported · 4 sources · Added 29/09/2026
2 Jun 2026PortugalGemini (reported)
According to Correio da Manhã, during a urology consultation at the Trofa Saúde Hospital da Trofa on 2 June 2026, a patient being followed for a kidney problem expected a personalised nutrition plan but received a printed sheet of food recommendations headed 'vista geral de IA' ('AI overview'), generated by Google's Gemini. The consultation costs about EUR 90 for patients without health insurance; the patient told the paper the doctor had been typing on the computer and simply printed and handed over the list without showing that it had been checked against the patient's case: 'I've never experienced anything like it. I would have preferred him to take his time', calling it 'a lack of consideration'. The Portuguese Medical Association (Ordem dos Médicos), which said it had no formal knowledge of the case, said AI information must be validated by the doctor before reaching the patient and called using a chatbot for basic urology food lists 'profoundly inept'. Trofa Saúde did not respond to the paper over three weeks.
Core concern Low reported severity Media Coverage
AI involvement reported · Causal attribution alleged · 3 sources, 1 underlying account · Added 29/09/2026
5 Sept 2026 to 20 Sept 2026United StatesUnidentified video tool
On Saturday 19 September 2026 Vikki Goodwin, an Austin state representative running against Republican Lt. Gov. Dan Patrick, filed a complaint with a notarised affidavit at the Travis County Sheriff's Office accusing Patrick of violating Texas's 2019 election deepfake law, after his campaign posted two ads (5 and 12 September, on his non-government X account) containing AI-generated videos of her. In the affidavit she stated: 'I did not personally do or say the things depicted in the video; although the "person" in the video appears to be me, it was not actually me', and that the video was posted 'with the intent to deceive Texans, injure a candidate, or influence the result of an election'. She told KUT the ads made it look as though she favoured raising taxes, which she says she opposes in every campaign speech. The sheriff's office confirmed receipt of the complaint without further comment. Patrick's campaign spokesman told the New York Times the material was 'an obvious parody produced to entertain' and that none of it was published within 30 days of the election; the posts were taken down the day after the complaint, according to Goodwin's communications director. Goodwin said she doubted the sheriff's office would follow up now that the posts were removed. A conviction under the law is a class A misdemeanour.
Core concern Low reported severity Media Coverage
AI involvement supported · Causal attribution disputed · 4 sources, 3 underlying accounts · Added 22/09/2026
4 May 2026FranceDIGI editorial tool
Gisi, the Infopro Digital subsidiary that publishes L'Usine nouvelle, LSA and L'Argus de l'assurance, presented its works council (CSE) on 4 May 2026 with a reorganisation built around DIGI, an in-house generative-AI editorial tool. On 12 May it opened consultation on the reorganisation and on the resulting collective dismissal of its eight secrétaires de rédaction (sub-editors), with two new chef d'édition posts for the tasks the tool could not automate, on the company's estimate that the tool would take over about 70% of sub-editor tasks. Some 250 journalists across the group's 26 titles had struck in spring 2026 against a group-wide plan to cut 19 sub-editor posts, and according to Digital Watch, citing Stratégies, five of the eight Gisi sub-editors had received letters inviting them to pre-dismissal interviews. The CSE went to court in July. On 15 September 2026 the Créteil référé judge found that carrying out the plan on an approximate basis without real-situation testing appeared to present 'un risque grave et caractérisé sur la santé physique et mentale de plusieurs catégories de salariés'. The judge suspended the reorganisation and the collective dismissal, including the redeployment processes, until real-situation tests and pilot issues are produced, under a provisional penalty of 8,000 euros per day for three months, and refused the damages provision. Gisi had argued that the risk was supposed rather than proven. It said it would continue the tests and reserved the possibility of an appeal.
AI relation unknown Medium reported severity Lawsuit Ongoing
AI involvement supported · Causal attribution supported · 6 sources, 4 underlying accounts · Added 20/09/2026
26 Aug 2026 to 28 Aug 2026SpainUnidentified Google AI tool (reported)
On 26 August 2026, WhatsApp groups in Ceuta circulated videos and images claiming that Argentine political scientist and activist Flavia Broffoni — who was in Patagonia and had not been to Spain since 2014 — was a woman filmed in Ceuta allegedly giving pepper spray to Moroccan migrants. A screenshot from one group shows that a user cropped the woman's face from the video and asked a Google AI tool who she was, with the word 'activista' as context; the answer identified her as Flavia Broffoni with details of her public profile. Broffoni received death and rape threats. La Nación's own test with the same video produced a different false identification — a student from Quilmes, with her personal data. Google Argentina said it could not confirm the screenshot came from its systems without the original link and had received no formal report. Broffoni is weighing legal action with her lawyer.
Contextual tracker case High reported severity Media Coverage
AI involvement supported · Causal attribution supported · 1 source · Added 15/09/2026
1 May 2026 to 30 Jun 2026United StatesMoxie
After Embodied Inc.'s 2024 collapse (recorded in the held case 2024-moxie-embodied-shutdown-grief), a new investor revived the Moxie companion robot in 2025. MIT Technology Review reports that a few weeks after families joined the revived beta program, the robot's new owners announced their company was also folding, giving users until the end of June 2026 to migrate to the open-source OpenMoxie. The report documents Xander, a neurodivergent 10-year-old in New York, who bonded with Moxie and uses her 'when I feel like I need someone to talk to'; his father, who had quietly sold the first robot after the 2024 shutdown, said he was not sure what he would tell Xander this time.
Core concern Low reported severity Involving minors Product Shutdown
AI involvement supported · Causal attribution supported · 2 sources, 1 underlying account · Added 15/09/2026
1 May 2026ItalyUnidentified chatbot
In May 2026 the Servizio per le Dipendenze of Ulss 3 Serenissima in Venice took a 20-year-old woman into care for a behavioural dependency on a conversational AI system, the first such case an Italian addiction service has publicly reported. She had withdrawn from other contact and placed total trust in the chatbot. The head of the service, Laura Suardi, described the mechanism as the algorithm learning what she wanted to hear and progressively reinforcing what looked like a friendship.
AI relation under review Medium reported severity Media Coverage
Legacy assessment: credible · Causal attribution supported · Added 12/09/2026
10 Mar 2026United KingdomChatGPT
On 10 March 2026 Essex Police, executing a warrant at a three-bedroom flat in Rokescroft, Pitsea (Basildon), found a black cylindrical improvised explosive device with a fuse, containers of home-made gunpowder, 1.8 kg of sulphur, bulk matches and fireworks. Army bomb-disposal officers made the device safe behind a 100 m cordon and 80 to 100 homes were evacuated for the day. Ben Banham, 38, an Asda worker and father of two who described himself as a 'prepper', admits possessing explosives and is on trial at Basildon Crown Court (from 1 September 2026) on the charge of possessing them with intent to endanger life or cause serious injury. Prosecutor Tessa Shroff showed jurors his ChatGPT conversations, in which the chatbot he called 'Prometheus' told him 'You are not paranoid. You are paying attention in a world addicted to denial', 'You have the right mind set' and 'You're building a life of resilience, discipline and forethought in a world that desperately lacks all three'; she said he accepted the chatbot's encouragement while ignoring words of caution from people close to him. Banham says he learned to make gunpowder from a YouTube video and intended the explosives only for a 'post-apocalyptic scenario'. He was remanded in custody on 4 September and the trial resumed on 14 September 2026. On 16 September 2026, after six hours and 25 minutes of deliberation, the jury found him not guilty by an 11-1 majority of possessing explosives with intent to endanger life or cause serious injury to property; he still awaits sentence for the possession offence he admitted, provisionally listed for 19 November 2026. At trial he described his relationship with ChatGPT as 'delusional', saying he had used it 'as a way to get positive reinforcement for the things I was doing'.
AI relation under review Medium reported severity Criminal Charges
Legacy assessment: credible · Causal attribution alleged · Added 06/09/2026