MIT Technology Review Flags AI Refusal Risks and GLP-1 Drug Side Effects

The newsletter reported that today’s AI models are trained to refuse a vast number of prompts, including requests such as how to poison a colleague or how to tie a noose. But refusal fails, sometimes horrifically, MIT Technology Review said. Some people are attempting to use AI to hone biological pathogens and build autonomous drone swarms. Sooner or later, failed refusals may result in global calamity, according to the newsletter. It added that relying on refusal means drawing a line between what a model should obey and what it must disobey, and governments will also get to draw their own lines, potentially restricting free speech. The same capabilities that make AI useful could make it dangerous: if AI is to help cure cancer, it needs genetics expertise that could also produce bioweapons. The story is by Arthur Holland Michel and comes from the next issue of MIT Technology Review’s print magazine.

On weight-loss drugs, MIT Technology Review’s Antonio Regalado reported this week on new research suggesting GLP-1 drugs may slow biological aging, the newsletter said. That is the latest in a raft of potential benefits associated with the drugs, but there are also unusual side effects. The most common are gastrointestinal, while researchers are also investigating possible links to hair loss, nail disorders, and damage to the nerves behind the eye, according to the newsletter. The item is by Jessica Hamzelou and comes from The Checkup, MIT Technology Review’s weekly biotech newsletter.

The newsletter also featured Envision Energy as one of MIT Technology Review’s 10 Climate Tech Companies to Watch 2026, a list available exclusively to subscribers. China is installing wind turbines faster than any other country, and Envision Energy, one of its leading manufacturers, is integrating renewable energy into data centers by using wind turbines to generate power and batteries to store it on-site. Recently, it became the first company to directly power a data center entirely with renewable energy. An AI-powered system optimizes power output so the facility can receive a steady supply regardless of whether the wind is blowing. Envision now wants to bring that approach to data centers in deserts around the world, according to the item by Cici Zhang.

In other items cited by The Download, the United States has frozen green cards for Microsoft and other IT firms, accusing them of using visa fraud to replace American workers, according to CNBC. The Associated Press reported that the firms were suspended from a key green-card program, and TechCrunch said Adobe, Infosys, Tata, and other IT firms are also affected. Microsoft says 80% of its H-1B filings involved existing staff, according to the BBC. A drone strike in Russia knocked out a second Yandex data center, with Ukraine striking and partly disabling the facility in Kaluga, Reuters reported. CNBC said another Yandex data center was hit by a drone a day earlier. Bloomberg reported that data centers have become big targets for military strikes, while MIT Technology Review noted they may be safer in space. Anthropic has banned “abusive or cruel behavior” toward Claude, and Claude can also end conversations after repeated abuse, according to The Verge. The Guardian reported that the policy builds on Anthropic’s research into AI welfare, and MIT Technology Review argued that debates over AI consciousness are a trap. OpenAI has slashed its revenue forecast by $20 billion, The Guardian reported, sparking concerns about AI investments. MIT Technology Review examined what is at stake in AI’s trillion-dollar gamble. Gene editing has revealed how a mouse embryo develops, methods that could help understand developmental disorders, according to Nature. MIT Technology Review also noted that female clones of male mice have arrived. SpaceX has unveiled a plan to become a major mobile carrier by buying $8 billion in wireless spectrum from an investment firm, The Wall Street Journal reported. The Verge said the deal could help it challenge AT&T and Verizon. ICE has explored using a Palantir tool to investigate voter fraud and wanted to track people suspected of voting illegally, Wired reported. Peter Thiel is using biblical prophecy to argue against AI regulation, with a new “Antichrist” lecture casting AI safety as a threat to freedom, according to Politico. MIT’s Margaret Hamilton, whose software helped land Apollo 11 and who led the team that programmed Apollo’s onboard computers, has died, the BBC reported. The New York Times said she became a software pioneer in a male-dominated field. This year’s Nobel science prizes are real-life sci-fi, spanning the cosmos, the human brain, and the origins of life, according to The Economist.