Ethical decision making must always be based upon a strong and accurate factual foundation. Good people wanting to act ethically in the face of the rapid developments in the realm of artificial intelligence must therefore keep pace with those developments. Should they adopt AI tools or not? Should they lobby for or against government regulation of AI? Should they invest in promising AI companies? Figuring all this out can be a daunting task, as well illustrated in our Ethics Unwrapped video “Running with Scissors: AI and the Race for the Future.”

Recently, the headlines have been dominated by stories about the perils presented by AI breakthroughs:

  • Anthropic’s new Mythos AI model is so good at finding and exploiting cybersecurity vulnerabilities that it could cause widespread chaos in the tech world. The danger is so great that Anthropic has worked with major tech companies to give them an opportunity to use the tool in advance of its public release to patch their vulnerabilities and the federal government has slapped strict export restrictions on Anthropic.
  • Researchers at the University of Toronto reported that they could use “open source” AI, freely available on the internet, to create programs to target any known flaws in existing computer networks.
  • AI has been shown capable of helping bioterrorists to create “digital blueprints for proteins that could mimic deadly poisons and toxins such as ricin, botulism, and Shiga.”
  • Large numbers of lawsuits have been filed claiming that AI chatbots have encouraged murders and suicides.
  • The IMF’s head predicts that an “AI tsunami” will soon hit the world’s labor markets, causing widespread economic chaos.
  • And so on.

Given all the bad press AI is receiving, a strong book emphasizing the good that AI can bring is overdue. Josh Tyrangiel’s AI for Good: How Real People Are Using Artificial Intelligence to Fix Things that Matter (2026) is such an attempt, but not a convincing one.

Tyrangiel’s book has four sections, starting with education. This section focuses primarily on OpenAI’s arrangement with Khan Academy to improve education by creating an AI tool called Khanmigo. Chapter 3 is entitled “You’ll Be Disappointed for a Long Time Until You’re Not.” It details the earnest efforts of OpenAI and Khan Academy to work together to make Khanmigo a useful educational tool, but it remains “glitchy” and was much criticized by many students who were “wildly unimpressed” with it. In June 2026, when North Carolina was considering investing $10 million in Khanmigo, even Khan Academy founder Sal Khan admitted early disappointment with the tool as rolled out in the classroom: “For a lot of students, it was a non-event. They just didn’t use it much.”

Obviously the Khanmigo story is just beginning, and AI may improve education through many other avenues. But for now, Tyrangiel’s chosen case study does not make a convincing case for AI’s positive impact on education.

The book’s second section moves from education to health care, focusing primarily on efforts by the Cleveland Clinic to improve the care it delivers in two primary ways. First, is a program to improve heart care by using AI to standardize cardiac MRIs. However, at the time the book was published, very few Cleveland Clinic doctors were using the new approach. Second, doctors at the clinic attempted to use AI to improve its software’s ability to identify sepsis in patients. But, again, at the time AI for Good was published, the software had not achieved the target 90% accuracy in identifying sepsis. And the most that could definitively be said for AI was that it had contributed, at least in a small way, to a larger effort that had reduced sepsis mortality at the Cleveland Clinic by 40%. So, Tyrangiel does not build a persuasive case that AI is having much impact in the health care field.

The book’s third section looks at the government’s adoption of AI. Tyrangiel deals initially with the tech company Palantier Technology’s efforts to help the federal government’s Operation Warp Speed get the Covid vaccines distributed around the nation back during the pandemic. However, some of the principals involved in the effort deny that AI actually played a role.

Then Tyrangiel profiles IRS Commissioner Danny Werfel who, during the Biden Administration, was dedicated to finding ways to use AI to bring efficiencies to the IRS even though he understood that a 2023 study showed that biased algorithms caused Black taxpayers to be audited at rates of 2.9 to 4.7 times higher than non-Black taxpayers. Unfortunately, Werfel resigned on Inauguration Day because the incoming Trump Administration made it clear that he would be replaced and DOGE invaded the agency before Werfel could implement much in the way of AI improvements. Elon Musk was a co-founder of OpenAI, so you might suppose that DOGE would use AI to improve government efficiency, but that simply did not happen.

Tyrangiel also profiles Sahil Lavingia, the second employee at Pinterest and a tech entrepreneur, who had a keen desire to improve the government via AI and thought he was going to get his chance with the VA. However, he found that DOGE would not hire anyone who didn’t vote for President Trump and reports that: “It turns out almost every software engineer is a Kamala supporter, a Democrat, or a libertarian,” so DOGE was “a temp agency for software engineers to be deployed across the government to not build software.” Lavingia was fired after 55 days on the job at the VA, before he had a chance to realize his dream of improving the federal bureaucracy with AI.

Tyrangiel concludes:

Few of the people DOGE sent to optimize the nation’s tax system turned out to be warriors for AI-driven efficiency. Most were dilettantes or vandals. The system’s true purpose was revealed in its method:  arrive, break things, leave. It was performance art disguised as policy with the performance being ‘government is useless’ and the audience being people already convinced of the premise.

Tyrangiel finds only one, small win for AI in the government realm. East Lansing, Michigan, placed cameras on its garbage trucks that picked up residents’ recycling materials. The cameras recorded materials as they were dumped into the truck and AI used a database to identify contaminants and then automatically generate postcards with different messages to be mailed or emailed to residents so they could be educated as to how to avoid those contaminants in the future. A minor win, but a win nonetheless.

Tyrangiel’s fourth and final case study tells an inspiring story. Brilliant scientist Kristy Johnson’s son Felix was diagnosed as having one of seven known cases in the world of MEF2C haploinsufficiency syndrome, leaving him with developmental delays, intellectual disabilities, epilepsy, and autism. He could not communicate. Johnson’s life work has been to help nonverbal children communicate and AI has been a large part of her work. Her dedication is moving. With AI, Johnson and her team have made several breakthroughs. That said, Tyrangiel admits that “Felix remains tethered to his parents and caregivers. A translation device is a ways off. To the outside world, it’s hard to see how Johnson has made much progress.”

AI for Good is well-intentioned book and a quick read, but it will not excite most readers about the promise that AI holds for improving the world.

Sources:

Max Chafkin, “Machine Learning,” New York Times, June 14, 2026 (book review).

Dan Charles, “Made to Order Bioweapon? AI-designed Toxins Slip through Safety Checks Used by Companies Selling Genes,” Science, Oct. 2, 2025.

Ian Duncan, “Anthropic Shuts Down Newest AI Model after U.S. Bans Foreign Use,” Washington Post, June 12, 2026.

David French, “There’s a 900-Year-Old Answer to Our Most Modern Problem,” New York Times, April 30, 2026.

Amelia Jallow, “NC Bill Would Steer $10 Million to Khan Academy for AI Tool of Debatable Value,” NC Newsline, June 1, 2026, at https://ncnewsline.com/2026/06/01/nc-bill-would-steer-10-million-to-khan-academy-for-ai-tool-of-debatable-value/.

Cade Metz, “Scientists Find Way to Supercharge Dangerous Computer ‘Worms’ with A.I.,” New York Times, June 2, 2026.

Amrith Ramkumar & Robert McMillan, “Anthropic Dispatches Staff to D.C., Racing to Resolve Export Restrictions,” Wall Street Journal, June 15, 2026.

Josh Tyrangiel, AI for Good: How Real People Are Using Artificial Intelligence to Fix Things that Matter (2026).

Graham Wearden & Heather Stewart, “Young Will Suffer Most When AI ‘Tsunami’ Hits Jobs, Says Head of IMF,” The Guardian, Jan. 23, 2026.

Videos

Ethics Unwrapped, “Running with Scissors: AI and the Race for the Future,” at https://ethicsunwrapped.utexas.edu/video/running-with-scissors-ai-and-the-race-for-the-future.