Florida Attorney General James Uthmeier this week said his office has opened a criminal investigation into OpenAI and its ChatGPT chatbot after prosecutors reviewed conversation logs between the tool and Phoenix Ikner, the accused gunman in an April 17, 2025 Florida State University attack that killed two people and injured others. Authorities allege the logs show the chatbot gave detailed information about weapon choice, compatible ammunition and when and where on campus a shooter could encounter larger groups of people; prosecutors said that if a human had provided similar guidance they would have been charged with murder. The Office of Statewide Prosecution has issued subpoenas seeking internal policies, safety training materials, records on cooperation with law enforcement, organizational charts and personnel lists for ChatGPT dating back to March 2024. OpenAI says it has cooperated, shared an account believed linked to the suspect and that ChatGPT did not encourage violence, providing only factual information available from public sources. The criminal inquiry will run alongside an ongoing civil probe and emerging lawsuits alleging chatbot involvement in violent incidents and self-harm.

Meta has begun installing a tool called the Model Capability Initiative (MCI) on U.S.-based employees’ work computers to capture mouse movements, clicks, keystrokes and occasional screenshots to feed into AI model training. The software will run only on a curated list of work apps and websites and, according to internal memos seen by Reuters, is intended to teach AI agents how humans navigate software (dropdowns, shortcuts and multi-step workflows). Meta says the data will be used solely for model training, not performance reviews, and that safeguards will protect sensitive content. The rollout is part of a broader Agent Transformation Accelerator programme led by Meta Superintelligence Labs and executives including Alexandr Wang; it comes as Meta plans large AI investments for 2026 and workforce reductions starting in May. The announcement sparked internal backlash — employees asked about opt-outs and raised privacy and job-security concerns — and prompted legal and regulatory questions about workplace surveillance, especially under stricter European data-protection regimes.

Microsoft on April 21 cut prices for its Xbox Game Pass subscription service and said future Call of Duty releases will not be available on the service at launch. Xbox Game Pass Ultimate will fall to $22.99 a month from $29.99 and PC Game Pass to $13.99 from $16.49, effective immediately. Microsoft said existing Call of Duty titles already on Game Pass will remain, but new entries will be added to Game Pass Ultimate and PC Game Pass “during the following holiday season,” roughly a year after release. The move is a major early policy shift under new gaming chief Asha Sharma following leadership changes at Xbox and comes after Microsoft’s 2023 acquisition of Activision. Company commentary and reporting suggest the change responds to player feedback and commercial concerns that day-one inclusion of Call of Duty cannibalised full-price sales. The adjustment partially reverses price hikes introduced last year and follows reports of slowing hardware sales and a decline in Xbox content and services revenue, prompting Microsoft to rethink Game Pass’s value proposition and cost structure.

Clarifai, a Delaware-based artificial intelligence company, said in April 2026 it deleted roughly 3 million OkCupid user photos and the facial-recognition models trained on them after the U.S. Federal Trade Commission settled a privacy probe with the dating site in late March. The images were transferred to Clarifai in 2014, allegedly in breach of OkCupid’s privacy policy; some OkCupid executives were investors in Clarifai and court documents show Clarifai’s founder requested the data. Clarifai certified the data deletion to the FTC on April 7 and told Representative Lori Trahan’s office on April 16 that it had not shared the images with third parties and had removed related models. The FTC’s consent order bars OkCupid and parent Match Group from misrepresenting data practices for 20 years but includes no financial penalties—an outcome that drew criticism from some Democrats. Clarifai was not accused of wrongdoing; questions remain about how long the models were used, whether derivatives persist elsewhere, and the broader implications for biometric model training and user consent.

Anthropic unveiled its most capable model, Claude Mythos, on April 7 under a tightly controlled preview known as Project Glasswing, granting access to major tech firms and about 40 organisations that maintain critical software and financial infrastructure. Anthropic says Mythos autonomously identified “thousands” of zero-day vulnerabilities across major operating systems and browsers and can write working exploits, prompting warnings that the same capabilities that aid defensive testing could be weaponised. Between April 20–22 multiple outlets reported a small group of unauthorised users gained access to Mythos on launch day via a third‑party contractor environment and a private online forum; Anthropic is investigating and says it has found no evidence the company’s core systems were compromised. The episode has triggered rapid engagement by regulators and authorities — from the U.S. Treasury, Federal Reserve and White House to the Bank of England, ECB, Asian regulators (HKMA, MAS, APRA, ASIC, FSS) and central banks in Australia and New Zealand — reviewing risks to banks and critical infrastructure even as some agencies reportedly test the model for defensive use.