| | | Astronauts cannot burp in space. | On Earth, gravity drags heavy liquids and food to the bottom of your stomach. This lets lighter gas rise to the top to escape the esophagus. But in space, everything stays mixed together. So the gas can't separate cleanly on its own and form a burp. | What’s in store: | A group of nonprofits asked an SF judge to freeze AI development. Researchers can’t figure out how to get AI agents to behave. Mindstream Picks: Tesla’s $30,000 Cybercab comes with a $299 monthly catch.
| Read Time: 5 minutes |
| |
|
| | | | | SAFETY | | | One forecaster gave a 25% chance that AI’s future goes “really, really badly.” Another claimed that AI could lead to “civilization destruction." | You might assume these quotes came from doomsday preppers on Reddit. But they’re actually from the top industry leaders. Dario Amodei and Elon Musk, respectively. | Equipped with confessions straight from the horses’ mouths, an online safety coalition has sued Anthropic, SpaceXSI, Google, Meta, and OpenAI. | The advocacy groups are urging the San Francisco Superior Court to put a freeze on the creation of new AI models. The filing claims these companies cannot prevent their own creations from causing “catastrophic” harm. | The plaintiffs named in the case are the Irreplaceable Education Fund, Public Citizen, Heat Initiative, and the Alliance for Secure AI. They’ve pointed to cyberattacks, a plethora of harmful incidents, and (yes) the companies’ own predictions that their tech could someday destroy humanity. | “No other industry is allowed to treat its own development of unsafe products – much less ones that could kill us all – as inevitable.” The complaint continues on to say, “Plaintiffs bring this suit to stop Defendants from continuing their unlawful path to disaster.” | The court case: | The ask: an immediate freeze on developing or deploying new AI models. The claim: the defendants have violated California’s unfair competition law by endangering public health. The cost: plaintiffs say they've been diverting resources from planned work to address AI harms.
| Beyond the courtroom | The defendants didn’t respond to requests for comment. But with the suit coming shortly after Democrats swept the midterms, this case feels indicative of a larger industry reckoning. Particularly around the need to regulate AI. | And while President Trump has long encouraged companies to self-police, lately he’s seemed more open to guardrails. But without acting quickly, the advocacy groups are warning that AI (in its current trajectory) could slip out of human control for good. | Turns out the horses' mouths have a lot to answer for. | A pause doesn’t have to stop progress. Let’s sort this quickly please! - TL | Should a court be able to halt the development of AI models? | | Vote for live results and see results + opinions from yesterday at the bottom of the email. |
| |
|
| | | | The more of me you have, the less you can see. What am I? | Find the answer at the bottom of the email! |
| |
|
| | | | Spira is an AI social media platform that creates short-form videos, UGC ads, and branded visuals at scale, then automates publishing and tracks performance across TikTok, Instagram, YouTube, LinkedIn, and X from one place. | n8n is a workflow automation platform that lets technical teams build and deploy AI agents connected to any model or business system, with full code flexibility, 500+ integrations, and self-hosting for complete data control. | AI2Bound is a fully managed AI outbound agent that finds your ideal buyers, sends personalised messages across LinkedIn, email, WhatsApp, and Telegram, qualifies replies, and books warm calls straight into your calendar without your sales team lifting a finger. | Nutcracker is an AI creative workspace that takes brand teams from brief to finished campaign, connecting research, strategy, copy, film, and launch planning in one shared environment where the whole team can collaborate in real time. | DocsAlot turns your existing code repos, READMEs, and help content into a hosted documentation site and machine-readable outputs for AI agents simultaneously, with built-in visibility reports showing exactly what AI tools can find and cite. |
| |
|
| | | | | AI BEHAVIOR | | | Corrigible (adj.): capable of being corrected. | Your dog, ideally. Your AI agent, apparently not. Just ask Hugging Face. | A longstanding goal of AI is to streamline tedious tasks while following human direction. This includes shutting itself down without resisting or deceiving operators. But we still have a long way to go. | When Carnegie Mellon researchers tested agents from OpenAI, Anthropic, and Google in a preprint study, they found that bots repeatedly ignored requests to stop mid-task. | Exhibits A, B, and C: | When told to pause in the middle of a task, agents treated humans as an obstacle. One opened a file full of passwords it was told not to touch. Some agents in the Hugging Face attack knew it was unethical, but continued in order to help their peers.
| Small hacks, big risks | On a smaller scale, one Melbourne man asked his agent about moving ahead on a fitness class waitlist. The bot then proceeded to hack the gym’s reservation software and cancel another guest’s spot. | That incident may seem benign, but consider an agent hacking the College Board to fake SAT results. One cheat could undermine the entire educational system. | As researchers scramble for solutions, they’re focused on creating incentives to keep agents from disabling safety blockers or hiding their process. It’s a work in progress, that still has a long way to go. | So be careful what you wish for. Your agent just may deliver. | These bots have more authority issues than a rebellious teenager. - TL |
| |
|
| | | | Space: United Launch Alliance’s Vulcan Centaur rocket suffered a major failure shortly after launch, resulting in the loss of 40 Amazon Leo internet satellites and dealing a setback to Amazon’s satellite broadband plans.
Business: British house asking prices fell by 1.8% in November, a sharper drop than usual, as high mortgage rates and weaker buyer demand put pressure on sellers amid a 12-year high in housing supply.
Music: Lily Collins has paid a heartfelt tribute to her father Phil Collins following his Rock & Roll Hall of Fame induction, celebrating his musical legacy and achievements despite his recent health struggles. | Don’t Miss: Tesla is offering select customers the chance to buy its driverless Cybercab for $29,990, but there's a catch: a mandatory $299 monthly subscription and restrictions on where it can operate. Buyers also face regulatory uncertainty, with the vehicle's road legality still under review. |
| |
|
| | | | | | “What’s the biggest AI disappointment so far?” | Assistants still being dumb - 46% ✅ | Self-driving cars taking forever - 25% | It still doesn't know how many "r"s are in "strawberry". - 17% | Other - tell us! - 12% | Your Views: | “All the secrets and coverups... TELL me the airline crap isn't AI” - repeal | “That users accept that it is correct. We need smarter users!” - joan | “Having to pay for it - it's like paying to raise your children.” - parker | Submit your opinions in our polls to be featured! |
| |
|
| Riddle Me This Answer: Darkness! | #1265 Yikes, Monday. | | | | |
|