Certain Chatbots Vastly Worse For AI Psychosis, Study Finds: “There’s no longer an excuse for releasing models that reinforce user delusions so readily.”
April 23. 2026: A new study argues that certain frontier chatbots are much more likely to inappropriately validate users’ delusional ideas — a result that the study’s authors say represents a “preventable” technological failure that could be curbed by design choices. “Delusional reinforcement by [large language models] is a preventable alignment failure.”
______________
Aprii 23. 2026: A handful of users from a private Discord channel gained unauthorized access to Mythos and have been using it regularly, not for cybersecurity purposes. Interestingly, users didn’t gain access through a sophisticated hack but by guessing the model’s online location based on past Anthropic naming conventions.
______________
Nvidia CEO Says AI Will Be a Permanent Micromanaging Boss Who Never Stops Nagging You
April 22, 2026: A recent panel at Stanford University, when Nvidia CEO Jensen Huang painted an unusual picture of an AI agent-dominated future… human workers’ productivity will instead go through the roof — with the minor tradeoff that ….“Your [AI] agents … harassing you, micromanaging you, and you’re busier than ever,” Huang said. “And yet our company is able to do more.”…, “we’re gonna create more jobs in the end,” he argued.
______________
Meta Installing Software on Employee Computers to Track Everything They Do, Feed the Data to AI
April 22. 2026: The company is saying the quiet part out loud… reportedly looking to develop AI agents that can complete work tasks autonomously, in perhaps one of the more conspicuous efforts to automate human workers’s jobs we’ve come across … forcing employees to train their AI.
______________
Anthropic reportedly ‘lost control’ of its most dangerous AI model — and that should worry everyone
April 22. 2026: Just last week, Anthropic launched… Claude Mythos, a model reportedly considered too dangerous for broad release. Now, the company is facing uncomfortable questions after reports claimed an unauthorized group gained access to Claude Mythos.
______________
Call it a glide path, or a slippery slope, but all the celebrities copping to their AI interests is changing the mood in Hollywood.
April 22. 2026: Natasha Lyonne… an actor who also writes, directs, and produces, has a reputation as an eccentric and intelligent cinephile—a “cool girl.” Not someone who would side with the machines over the artists. “Suddenly I became some weird Darth Vader character or something,” Lyonne told Variety of the backlash. “It’s not fun when people say not nice things to you.
______________
Google is doing just fine on AI
April 22. 2026: Google is facing a new wave of heat from critics who say it’s falling behind on AI coding and that even its own employees are choosing Claude Code over its homegrown products…. it’s wrong now…. Google is stepping right in to gobble up everyday users that are already accustomed to turning to the tech company to run their email, meetings, and other large swaths of their lives.
______________
Chinese Workers Horrified as Bosses Direct Them to Train Their AI Replacements: Why would they choose to dig their own graves?
April 21, 2026: For years, a buzzy Silicon Valley startup called Mercor has been hiring an army of desperate job-seekers — often including educated and underemployed experts — to train AI models to replace them in the workforce…. the business world continues to push for automation, hoping to wean itself off relying on pesky and expensive human labor once and for all.
______________
New AI tool can predict heart failure five years early with 86% accuracy; “This will allow doctors to make more informed decisions about the best way to treat patients.”
April 15, 2026: Some 350,000 tests are ordered in the United Kingdom each year for chest pain — and AI, if used — can predict if the patient is at high risk of experiencing heart failure five years early. (See also ”deep-learning model to forecast a patient’s heart failure.”)
###
About The Author
