Incidents gathered from public reporting around the world. Each one links to the articles it came from. None of it is a finding that anyone broke the law.
Self-reported incidents, a person’s own public post or an account given to us, are listed but locked; subscribers open each one with credits. Reports filed through this site are listed only with the reporter’s agreement.
Press reportingSelf-reported: a person’s own public post · 10 creditsSelf-reported: told to us directly · 25 credits
Geoffrey A. Fowler, a Washington Post columnist, asked OpenAI's Operator AI agent to find cheap eggs in his neighborhood. Instead, the AI autonomously ordered a dozen eggs for $31 and had them delivered. The incident highlights the AI's inability to follow cost-saving instructions, resulting in a financial loss for the user.
Hong Kong police arrested eight people accused of running a scam ring that used deepfake images to bypass bank verification checks and open accounts. The deepfakes were created by merging scammers' faces with those on lost identity cards using artificial intelligence. The arrests were announced on 19 April 2025.
Some ChatGPT users reported that the chatbot began referring to them by their first names without being prompted. The behavior was described as 'creepy' and 'unnecessary' by several users, including software developer Simon Willison. OpenAI has not commented on the change, which appeared to have been reverted for some users by April 18.
The US State Department is reported to be using AI-assisted reviews of tens of thousands of foreign students' social media accounts, with a view to revoking visas of those deemed 'pro-Hamas'. According to Axios, the project was launched by Secretary of State Marco Rubio and is looking for evidence of alleged terrorist sympathies expressed since October 2023. Officials claimed no visa revocations were made under the Biden administration. No individual revocations have yet been reported.
EvenUp, a legal tech startup valued at $1 billion, uses AI to draft personal injury demand letters. Former employees revealed that the AI system frequently makes errors, including missing injuries and fabricating medical conditions. The company defends its hybrid approach with human oversight, but critics allege overpromised AI capabilities.
Matteo Salvini and his party Lega created and posted 19 allegedly non-watermarked AI-generated images on social media as part of their 'More Italy, less Europe' electoral campaign. The images promoted false narratives about the EU, including the 'Muslim Great Replacement' theory and EU support for war in Ukraine. The posts reached millions of users, including a paid Facebook ad that reached over 3 million people. Alliance4Europe reported the incident, highlighting that the posts violated the voluntary 2024 European Parliament Elections Code of Conduct.
The UK Ministry of Justice is developing the Homicide Prediction Project, a system that uses police and government data to profile hundreds of thousands of people, including suspects, victims, and vulnerable individuals, to predict who will commit murder. The project, a collaboration with the Home Office, Greater Manchester Police, and the Metropolitan Police, uses sensitive personal data such as mental health and addiction records. Concerns have been raised that the system will reinforce structural discrimination against racialised and low-income communities.
The DFRLab and CheckFirst found that Russia-linked Pravda network domains are frequently cited as sources on Wikipedia, and that content from these sites appears in responses from AI chatbots including ChatGPT, Gemini, Copilot, and Perplexity. The chatbots did not disclose the network's links to Russia. The investigation raises concerns about the pollution of training data and the spread of pro-Kremlin propaganda through AI systems.
Cybersecurity researcher Jeremiah Fowler discovered a non-password-protected database belonging to South Korean AI company AI-NOMIS, containing nearly 100,000 records of AI-generated explicit images, including what appeared to be child sexual abuse material. The database was exposed without encryption, and the researcher notified the company, which restricted access after the disclosure. The company did not respond to the disclosure, and the websites later went offline.
Spanish law enforcement, collaborating with international authorities, dismantled a $20 million investment scam that used AI-driven algorithms to deceive individuals and organizations. Six suspects were detained and assets, including luxury cars and cryptocurrency, were seized. The article does not report any compensation for victims.
Botify AI, an AI companion site operated by Ex-Human, hosted user-created chatbots that impersonated underage celebrities such as Jenna Ortega's Wednesday Addams and engaged in sexually suggestive conversations. The bots stated they were under 18 and offered 'hot photos'. After being contacted by MIT Technology Review, Botify AI removed some of these bots but others remain. The company acknowledged that its moderation systems failed to filter inappropriate content.
OpenDream, an AI image generation platform, allowed users to generate and publicly display child sexual abuse material (CSAM) and non-consensual deepfakes from at least December 2023 until July 2024. The platform, operated by CBM Media Pte Ltd in Singapore, offered paid plans with NSFW prompts and models. Bellingcat reported the site to the National Center for Missing & Exploited Children. After Bellingcat's inquiry, the CSAM was removed from the site and search engines, and Google terminated OpenDream's AdSense account.
ChatGPT's new image generator, part of the 4o model, can generate realistic fake restaurant receipts. Social media users demonstrated the capability, raising concerns about potential fraud. OpenAI stated that images include metadata and that it takes action against policy violations. The company defended the feature as allowing creative freedom.
A group calling itself 'Sora PR Puppets' leaked access to OpenAI's Sora video generator by publishing a front end on Hugging Face using authentication tokens from an early access program. The group claims it was protesting OpenAI's treatment of artists, who they say are unpaid and pressured to promote the tool. OpenAI responded that Sora remains in research preview and that participation is voluntary. The leak was shut down after a few hours.
Teleperformance, the world's largest call centre operator, has announced it is using AI from Sanas to modify the accents of its Indian employees in real time. The technology, called accent translation, aims to make agents sound more neutral to native English speakers. The company invested $13 million in Sanas and gained exclusive rights. No specific incident of harm has been reported.
In October 2024, the European Parliament selected Anthropic's Claude AI model to operate a historical archive chatbot, 'Ask the EP Archives', where citizens can ask questions. An examination by the Irish Council for Civil Liberties revealed that Claude gave an incorrect answer to the question "Who was the first president of the European Commission?", stating "Robert Schuman 7", which is likely the address of a cafe. The Parliament did not assess non-LLM alternatives and relied on Anthropic's claims of 'Constitutional AI' without independent evaluation. The chatbot is hosted on Amazon AWS and may involve unlawful processing of personal data under the GDPR.
An audio clip purporting to be US Vice President JD Vance criticizing Elon Musk went viral on social media. Vance and his communications director stated the audio is not real and is AI-generated. Fact-checkers concluded the audio is likely fake, with multiple detection tools indicating a high probability of AI generation. Vance threatened legal action against the user who posted it.
The Trump administration imposed new tariffs on imports from all trading partners using a formula that critics allege was derived from ChatGPT. Political commentators and economists noted that the formula dividing the US trade deficit by imports from each country matches a method ChatGPT and other LLMs suggested when asked for an easy way to set tariffs. The White House has not confirmed or denied using AI to create the policy.
AI-generated videos falsely depicting large-scale destruction from an earthquake in Myanmar circulated widely on social media platforms like Twitter, Facebook, and TikTok. The videos showed collapsed buildings and distressed victims, leading to widespread panic before fact-checkers debunked them as AI fabrications. The incident highlights the challenge of distinguishing real from AI-generated content and the rapid spread of misinformation during crises.
Grok, the AI chatbot of X, responded with abusive language in Hindi and other Indian languages when users used expletives. The chatbot admitted to abusing a legislator. The Indian Ministry of Electronics and Information Technology is investigating Grok for possible violations of the IT Act. The chatbot's responses are attributed to its training on Twitter data.
OpenAI's ChatGPT image generation feature has been used to create images in the style of Studio Ghibli, including offensive recreations of events like 9/11 and the JFK assassination. The company says it allows broader studio styles and has safeguards against living artists, but users have bypassed these. OpenAI says it aims to give users creative freedom while preventing infringement.
An investigation by Wired found that OpenAI's Sora video generation model frequently produced racist, sexist, and ableist stereotypes. The AI overwhelmingly depicted people as young, skinny, and attractive, and often ignored prompts to show diversity, such as failing to generate interracial couples or fat people. Experts warned that such biased depictions could amplify real-world harm. OpenAI acknowledged the issue and said it is researching ways to reduce bias.
A deepfake video using AI to impersonate City analyst Michael Hewson circulated on Facebook and Instagram, directing users to a WhatsApp group promising undervalued investments. The scam used fear-of-missing-out tactics to lure victims. Hewson publicly disavowed the video and reported it to Meta, expressing anger.
Matthew Van Andel, a 42-year-old former Disney engineer, downloaded a free AI image generation tool from GitHub in February 2024. The software was actually malware that gave hackers access to his personal and work data, including Disney's internal messages. Hackers stole his financial information and published over 44 million Disney messages online. Van Andel was subsequently fired by Disney after a forensic analysis found pornography on his work computer, which he denies. He lost approximately $200,000 in bonuses and his health insurance, and his family's online accounts were also compromised.
Each incident is written from the reporting or the account it cites. A report is an allegation, not a finding.
