The UK's Information Commissioner's Office opened a formal investigation on February 3, 2026 into X.AI and X over whether Grok unlawfully processed personal data to generate non-consensual sexualised images, including of children.
The FTC finalized a $930,000 settlement on August 28, 2026 with Cox Media Group and two partners after finding their 'Active Listening' AI ad product never actually collected voice data and instead resold brokered email lists.
Florida Attorney General James Uthmeier sued OpenAI and Sam Altman personally on June 2, 2026, seeking a court order to stop ChatGPT from collecting data from children under 13 without parental consent.
Kentucky Attorney General Russell Coleman sued Character.AI on January 12, 2026 under the state's brand-new Consumer Data Protection Act, alleging the chatbot collected children's data without parental consent and used it to train its AI models.
Voice-biomarker startup Kintsugi shut down in February 2026 after spending $30 million over seven years, including roughly $16 million on FDA pre-submission work, and never filed its De Novo application despite a pivotal study showing its AI matched clinical depression screening accuracy.
A Stanford-led study of 1,131 Character.AI users published in Nature Human Behaviour found that people with small social networks who leaned on chatbots for companionship reported lower well-being, and that sharing more with the AI was linked to feeling worse, not better.
A UNC Chapel Hill study of more than 2,300 middle schoolers found AI chatbot use jumped from 62% to 75% in four months, and students who used chatbots for emotional companionship went on to report greater loneliness and higher social anxiety.
Dartmouth's Therabot chatbot cut depression symptoms 51% in a 210-person randomized trial, and its creators are now racing to commercialize it even though the FDA has yet to clear any AI device specifically for mental health.
The FDA granted Breakthrough Device Designation in March 2026 to PathAI's PathAssist Derm, an AI tool meant to help pathologists on ambiguous skin biopsies — a category where studies show diagnostic concordance between pathologists falls below 50%.
A deep learning model called AgeNet, trained on over 420,000 chest X-rays from Korean adults and tracked for a median of 8.5 years, found that patients whose AI-estimated 'radiographic age' outpaced their real age had up to 52% higher mortality risk, per a study published in Radiology: Artificial Intelligence.