For decades, handmade narco subs have been some of the cocaine trade's most elusive and productive workhorses, ferrying multi-ton loads of illicit drugs from Colombian estuaries toward markets in North America and, increasingly, the rest of the world. Now off-the-shelf technology--Starlink terminals, plug-and-play nautical autopilots, high-resolution video cameras--may be advancing that cat-and-mouse game into a new phase. Uncrewed subs could move more cocaine over longer distances, and they wouldn't put human smugglers at risk of capture. And law enforcement around the world is just beginning to grapple with what this means for the future. This story is from the next print issue of magazine, which is all about crime. Google DeepMind is calling for the moral behavior of large language models--such as what they do when called on to act as companions, therapists, medical advisors, and so on--to be scrutinized with the same kind of rigor as their ability to code or do math.

Russian attacks on Ukraine energy sites'particularly depraved', UK PM Starmer says Russia's attacks on Ukraine's energy sector on Monday night - as temperatures dropped to -20C (-4F) - were barbaric and particularly depraved, UK Prime Minister Sir Keir Starmer has said. He made the comments after speaking to US President Donald Trump hours after Russia hit power plants and critical infrastructure in the capital, Kyiv, and elsewhere. The attacks came at the end of a week-long pause that Trump had asked Russia's President Vladimir Putin to observe as a fierce cold swept Ukraine. Trump said on Tuesday that Putin had kept his word and that he would like him to end the war. Top US envoys are meeting negotiators from Russia and Ukraine in Abu Dhabi on Wednesday and Thursday.

Russia has launched its most powerful blow against Ukraine's energy sector so far this year, according to the private energy company, DTEK. The combined missile and drone strikes which targeted power plants and infrastructure in Kyiv and multiple locations left the system operating with serious restrictions, it said. The strikes were launched as temperatures dropped to -20C (-4F) and left more than 1,000 tower blocks in the capital without heating once again and damaged a power plant in the eastern city of Kharkiv beyond repair. President Volodymyr Zelensky said Russia was choosing terror and escalation rather than diplomacy to end this war and called for maximum pressure on Moscow from Ukraine's allies. The attack comes after a so-called energy truce agreed by Donald Trump with Vladimir Putin expired at the weekend.

US support for nuclear energy is soaring. Meanwhile, coal plants are on their way out and electricity-sucking data centers are meeting huge pushback. Welcome to the next front in the energy battle. Take yourself back to 2017. Get Out and The Shape of Water were playing in theaters, Zohran Mamdani was still known as rapper Young Cardamom, and the Trump administration, freshly in power, was eager to prop up its favored energy sources. That year, the administration introduced a series of subsidies for struggling coal-fired power plants and nuclear power plants, which were facing increasing price pressures from gas and cheap renewables.

If a major disaster like Fukushima or Chernobyl ever happens again, the world would know almost straight away, thanks to an array of government and DIY radiation-monitoring programs running globally.

Zanskar used AI tools to identify a site that could host a commercial power plant. Zanskar used AI tools to help revive a New Mexico geothermal plant. Now, the company found a hotspot that could support a new power plant. Sometimes geothermal hot spots are obvious, marked by geysers and hot springs on the planet's surface. But in other places, they're obscured thousands of feet underground. Now AI could help uncover these hidden pockets of potential power.

Zanskar uses AI to identify hidden geothermal systems--and claims it has found one that could fuel a power plant, the first such discovery by industry in decades. A geothermal startup said Thursday that it has hit gold in Nevada--metaphorically speaking. Zanskar, which uses AI to find hidden geothermal resources deep underground, says that it has identified a new commercially viable site for a potential power plant. The discovery, the company claims, is the first of its kind made by the industry in decades. The find is the culmination of years of research on how to find these resources--and points to the growing promise of geothermal energy .

As Large Language Models become increasingly integral to various domains, their potential to generate harmful responses has prompted significant societal and regulatory concerns. In response, governments have issued ethics guidelines to promote the development of trustworthy AI. However, these guidelines are typically high-level demands for developers and testers, leaving a gap in translating them into actionable testing questions to verify LLM compliance. To address this challenge, we introduce GUARD (\textbf{G}uideline \textbf{U}pholding Test through \textbf{A}daptive \textbf{R}ole-play and Jailbreak \textbf{D}iagnostics), a testing method designed to operationalize guidelines into specific guideline-violating questions that assess LLM adherence. To implement this, GUARD uses automated generation of guideline-violating questions based on government-issued guidelines, thereby testing whether responses comply with these guidelines. When responses directly violate guidelines, GUARD reports inconsistencies. Furthermore, for responses that do not directly violate guidelines, GUARD integrates the concept of ``jailbreaks'' to diagnostics, named GUARD-JD, which creates scenarios that provoke unethical or guideline-violating responses, effectively identifying potential scenarios that could bypass built-in safety mechanisms. Our method finally culminates in a compliance report, delineating the extent of adherence and highlighting any violations. We have empirically validated the effectiveness of GUARD on seven LLMs, including Vicuna-13B, LongChat-7B, Llama2-7B, Llama-3-8B, GPT-3.5, GPT-4, GPT-4o, and Claude-3.7, by testing compliance under three government-issued guidelines and conducting jailbreak diagnostics. Additionally, GUARD-JD can transfer jailbreak diagnostics to vision-language models, demonstrating its usage in promoting reliable LLM-based applications.

This week, Casey Crownhart, senior reporter for energy at MIT Technology Review and Pilita Clark, FT's columnist, consider how China's rapid renewables buildout could help it leapfrog on AI progress. In the age of AI, the biggest barrier to progress isn't money but energy . That should be particularly worrying here in the US, where massive data centers are waiting to come online, and it doesn't look as if the country will build the steady power supply or infrastructure needed to serve them all. For about a decade before 2020, data centers were able to offset increased demand with efficiency improvements . Now, though, electricity demand is ticking up in the US, with billions of queries to popular AI models each day--and efficiency gains aren't keeping pace. With too little new power capacity coming online, the strain is starting to show: Electricity bills are ballooning for people who live in places where data centers place a growing load on the grid.

Plus: Huntington's disease has been treated successfully for the first time. Wikipedia is the most ambitious multilingual project after the Bible: There are editions in over 340 languages, and a further 400 even more obscure ones are being developed. But many of these smaller editions are being swamped with AI-translated content. Volunteers working on four African languages, for instance, estimated to that between 40% and 60% of articles in their Wikipedia editions were uncorrected machine translations. This is beginning to cause a wicked problem. AI systems learn new languages by scraping huge quantities of text from the internet.