The Download: autonomous narco submarines, and virtue signaling chatbots

Uncrewed narco subs could transform the Colombian drug trade

Off-the-shelf technology is poised to revolutionize the cocaine trade. New, uncrewed narco submarines could move more drugs over longer distances without risking human smugglers.

For decades, handmade subs have been the trade's elusive workhorses. Now, Starlink terminals, plug-and-play autopilots, and high-resolution cameras are advancing the cat-and-mouse game with global law enforcement.

Google DeepMind scrutinizes chatbot morality

Google DeepMind is calling for the moral behavior of large language models to be rigorously tested. The company argues that as LLMs take on sensitive roles, their trustworthiness must be scrutinized like any other capability.

People are increasingly asking chatbots to act as companions, therapists, and medical advisors. With agents starting to take actions on behalf of users, the need to understand their real-world ethical alignment is urgent.

• LLMs may influence human decision-making.
• Their performance in roles like coding or math is tested, but their moral reasoning is not.
• No one knows how trustworthy this technology truly is in sensitive applications.

The legal case for global climate justice

The legal argument for holding major polluters financially responsible for climate damage is gaining strength. The United States and the European Union became economic superpowers by burning a disproportionate share of the world's oil and gas.

Morally, the case for compensation from responsible countries and companies has long been clear. Legally, it has been difficult to establish—but that may now be changing as new legal strategies emerge.

The must-reads

From a US censorship portal to AI's power grid, here are today's essential tech stories.

The US is building a portal to access banned content. The freedom.gov site is Washington's broad solution to global censorship. The Trump administration is also training an elite cadre of coders.

Mark Zuckerberg overruled experts on beauty filters. The Meta CEO reportedly kept the filters on Instagram, citing "free expression." This comes as a new investigation finds Meta is failing to protect children from predators on its platforms.

Silicon Valley is building a shadow power grid. AI firms are planning private power plants for US data centers. They are pushing a narrative that generative AI will save the planet, but critics demand better sustainability metrics.

Russian forces are struggling with Starlink and Telegram crackdowns. New restrictions have left troops without reliable means to communicate on the front lines.

Quote of the day

“Too often, those victims have been left to fight alone…That is not justice. It is failure.”

—UK Prime Minister Keir Starmer, announcing plans to force tech firms to remove deepfake nudes and revenge porn within 48 hours or risk being blocked in the UK.

Could AI make end-of-life decisions?

Bioethicists are developing an AI tool to help predict what incapacitated patients would want for their end-of-life care. The goal is to assist distressed family surrogates who must make these difficult calls.

David Wendler, a bioethicist at the US National Institutes of Health, hopes to start building the tool soon. However, critics raise ethical questions about training such a system and whether life-or-death decisions should ever be entrusted to AI.

We can still have nice things

A bit of comfort, fun, and distraction to brighten your day.

• The Oakland Library keeps a public log of weird artefacts found in books.
• A guide on how to keep temperamental orchids alive.
• New York's Transit Museum is holding a "Pizza Rat Debunked" event.
• Why mid-level British indie bands are treated like royalty in China.