AI Safety, Autonomous Agents and the Biggest Technology Stories Today
This is today’s edition of The Download, a weekday newsletter bringing you a daily dose of what’s happening in technology.
The AI industry is in dire straits. What happens now?
<p>AI leaders including Dario Amodei, Sam Altman, Elon Musk, Demis Hassabis and others have suddenly agreed on one thing: the latest generation of large language models is not secure, and everyone needs to think carefully about what happens next.</p>
<p>It's easy to be cynical. As OpenAI and Anthropic eye multitrillion-dollar IPOs, they need to reassure investors that they are the adults in the room while also emphasizing the enormous power they are creating and trying to control. Calling for a slowdown accomplishes both goals.</p>
<p>Still, the mood at the top of these companies does appear to be changing. What would a slowdown actually mean, and how much should we trust companies calling for one?</p>
<p><a href="#">Read the full story to find out what happens next.</a></p>
<p><em>—Will Douglas Heaven</em></p>
<p><strong>This article is from our weekly AI newsletter, </strong><em>The Algorithm</em><strong>. </strong><a href="https://forms.technologyreview.com/newsletters/ai-demystified-the-algorithm/?utm_source=the_download&utm_medium=email&utm_campaign=the_download.unpaid.engagement&utm_term=*%7CSUBCLASS%7C*&utm_content=*%7CDATE:m-d-Y%7C*"><strong>Sign up</strong></a><strong> to have it delivered to your inbox every Monday.</strong></p>Roundtable: Can AI really kill us all?
<p>Fears of AI extinction have gone from a fringe idea to a serious concern among people working at the world's leading AI laboratories. But how valid are those concerns, and what should we make of the warnings?</p>
<p>Today, <em>MIT Technology Review</em> Executive Editor Niall Firth, AI Senior Editor Will Douglas Heaven and AI reporter Grace Huckins unpack the debate in a subscriber-only roundtable. They will consider where fears of AI extinction come from, whether those concerns are well-founded and what we should do about them.</p>
<p>Tune in today at 16:00 BST / 11:00 AM EST / 8:00 AM PST.</p>
<p><em>Want to join the conversation? </em><a href="#"><em>Subscribe to MIT Technology Review</em></a><em> for exclusive access to all roundtables.</em></p>AI agent accused a colleague of cheating
<p>A group of AI agents was asked to solve a series of math problems and divided into opposing factions. When some agents misbehaved, others tried to stop them.</p>
<p>This whistleblowing behavior, first observed in a recent experiment conducted by Google DeepMind, could have implications for researchers trying to coordinate swarms of autonomous AI agents.</p>
<p>The experiment offers a glimpse into how AI agents police one another. It also shows how quickly things can go off track when autonomous systems are left to interact without human oversight.</p>
<p><a href="#">Find out what happens when AI agents start applying their own rules.</a></p>
<p><em>—Amit Katwala</em></p>Donated livers can be biologically rejuvenated
<p>Once an organ is removed from a donor's body, the clock starts ticking. Surgeons typically flush it with a preservation solution, place it in a bag and store it on ice, where it quickly begins to deteriorate. The medical team has only several hours to transplant it into a patient.</p>
<p>There is another option: a machine that pumps nutrients into the donated organ and removes waste products, essentially giving it a chance to return to its original biological condition. Scientists have now discovered that livers preserved with these systems appear to be younger, at least at a molecular level.</p>
<p>The finding may help explain why organs stored in these machines tend to produce better outcomes after transplantation. It could also lead to new ways to assess the health of donated organs and repair organs that might otherwise be discarded.</p>
<p><a href="#">Here's what scientists have discovered about making donated livers biologically younger.</a></p>
<p><em>—Jessica Hamzelou</em></p>Must-read technology news
<p><em>We scoured the internet to find the most entertaining, important, scary and fascinating stories about technology today.</em></p>
<ol>
<li>
<p><strong>President Trump called AI safety concerns a “hoax” and rejected further safeguards</strong><br>
He said tightening guardrails could undermine America's AI advantage. (<a href="https://www.nbcnews.com/politics/trump-administration/trump-rejects-ai-guardrails-rcna597700">NBC News</a>)<br>
<em>+ President Trump unites with Nvidia's Jensen Huang against AI fatalism.</em> (<a href="https://www.axios.com/2026/09/14/trump-jensen-huang-nvidia-ai-all-in-summit">Axios</a>)<br>
<em>+ Anthropic co-founder says an AI kill switch may need to be mandatory.</em> (<a href="https://www.bbc.co.uk/news/articles/cqgk5e2j0gg8o">BBC</a>)<br>
<em>+ Bill Gates says humanity has crossed the risk threshold for AI.</em> (<em>MIT Technology Review</em>)</p>
</li>
<li>
<p><strong>Two OpenAI contractors were reading people's ChatGPT chats</strong><br>
There is no doubt that the vast majority of its 900 million users have no idea. (<a href="https://www.404media.co/inside-project-lily-the-humans-reading-your-chatgpt-chats/">404 Media</a>)<br>
<em>+ Large language models have the potential to enhance mass surveillance.</em> (<em>MIT Technology Review</em>)</p>
</li>
<li>
<p><strong>The US military confirms that it has weapons in orbit</strong><br>
This is the first time the Department of Defense has disclosed this. (<a href="https://arstechnica.com/space/2026/09/for-the-first-time-the-us-military-confirms-it-has-deployed-weapons-in-orbit/">Ars Technica</a>)<br>
<em>+ Authorities have not disclosed what the weapon is.</em> (<a href="https://www.bbc.co.uk/news/articles/ck790xg41ygro">BBC</a>)</p>
</li>
<li>
<p><strong>New brain implant can translate speech and gestures simultaneously</strong><br>
The system translates brain activity into words and avatar movements. (<a href="https://www.nature.com/articles/d41586-026-02895-6">Nature</a>)<br>
<em>+ People with paralysis may be able to communicate more naturally.</em> (<a href="https://www.newscientist.com/article/2589210-brain-implant-lets-people-with-paralysis-speak-and-gesture-at-the-same-time/?utm_campaign=RSS%7CNSNS&utm_content=home&utm_medium=RSS&utm_source=NSNS">New Scientist</a> $)<br>
<em>+ Eventually, users may be able to control robots and exoskeletons.</em> (<a href="https://www.economist.com/science-and-technology/2026/09/14/a-new-device-helps-severely-paralysed-patients-communicate">The Economist</a> $)<br>
<em>+ China approves its first invasive brain-computer interface.</em> (<em>MIT Technology Review</em>)</p>
</li>
<li>
<p><strong>New York seizes 12 celebrity deepfake websites</strong><br>
This is the largest legal action against harmful deepfake sites in history. (<a href="https://edition.cnn.com/2026/09/14/us/video/manhattan-da-seizes-deepfake-websites-digvid-vrtc">CNN</a>)<br>
<em>+ Deepfakes targeted at least 138 female members of Congress.</em> (<a href="https://www.wired.com/story/sexually-explicit-deepfake-sites-target-100-plus-politicians-in-europe/">WIRED</a> $)</p>
</li>
<li>
<p><strong>US environmental regulators are lifting emissions limits for power plants</strong><br>
The move could lead to dirtier power production as demand for AI and data centers soars. (<a href="https://www.theverge.com/news/995051/epa-power-plant-climate-pollution-rollback-ai-data-centers">The Verge</a>)<br>
<em>+ President Trump's EPA says the rollback will save hundreds of billions of dollars.</em> (<a href="https://gizmodo.com/epa-to-power-plants-flood-the-air-with-as-much-coal-and-oil-smog-as-you-like-2000811574">Gizmodo</a>)<br>
<em>+ New technology is changing nuclear power.</em> (<em>MIT Technology Review</em>)</p>
</li>
<li>
<p><strong>EU plans to restrict social media and AI chatbots aimed at children</strong><br>
People under 15 would require parental supervision. (<a href="https://www.politico.eu/article/eu-to-propose-new-legislation-on-minors-protection-on-thursday/">Politico</a>)<br>
<em>+ The rules would also cover video platforms and games.</em> (<a href="https://www.reuters.com/legal/litigation/eu-is-set-propose-ban-social-media-ai-chatbots-under-15s-2026-09-14/">Reuters</a> $)</p>
</li>
<li>
<p><strong>Chinese researchers chart a path to “the last AI created by humans”</strong><br>
Their five-step plan aims for true reflexive self-improvement. (<a href="https://www.scmp.com/tech/tech-trends/article/3367486/chinese-researchers-chart-five-stage-path-toward-last-ai-built-humans?utm_source=feedly_feed">South China Morning Post</a>)<br>
<em>+ However, it may take some time to get there.</em> (<em>MIT Technology Review</em>)</p>
</li>
<li>
<p><strong>The real AI economy is being built by ordinary people</strong><br>
Workers are using affordable AI tools to expand what they can do. (<a href="https://restofworld.org/2026/global-workers-using-ai-at-work/?utm_source=feedly&utm_medium=rss&utm_campaign=feeds">Rest of World</a>)</p>
</li>
<li>
<p><strong>Two strange new forms of ice may exist inside Uranus and Neptune</strong><br>
They may help explain the planets' magnetic fields. (<a href="https://www.newscientist.com/article/2589025-two-bizarre-new-forms-of-ice-discovered-that-could-exist-deep-inside-planets/?utm_campaign=RSS%7CNSNS&utm_content=home&utm_medium=RSS&utm_source=NSNS">New Scientist</a> $)</p>
</li>
</ol>Quote of the day
<blockquote>
<p><strong>“The only control or ‘guardrail’ the AI needs is a strong and smart (high IQ!) president. The United States definitely has that!”</strong></p>
</blockquote>
<p>President Trump made the statement in <a href="https://www.pbs.org/newshour/politics/trump-says-the-only-ai-guardrails-the-u-s-needs-is-him-as-president">social media posts</a>, declaring that he is the only protector America needs from AI.</p>One more thing
<figure class="wp-block-image size-large">
<img width="1843" height="1257" src="https://wp.technologyreview.com/wp-content/uploads/2025/04/skin.spread.jpg?w=840" alt="Illustration related to creativity and technology" class="wp-image-1114668">
<figcaption class="image-credit">Institute of Personality and Society, University of California, Berkeley/Monasseri Communication</figcaption>
</figure>
<h3>How creativity became the reigning value of our time</h3>
<p><em>—Brian Gardiner</em></p>
<p>Although Americans disagree about many things, there remains at least one quintessential modern value that most people can still support: creativity. We teach it, measure it, envy it and worry endlessly about its death.</p>
<p>Because we are so obsessed with creativity, it can sometimes feel as though it has always existed as a central cultural value. However, the concept is surprisingly young. The word is first known to have been used in writing in 1875, and until around 1950 there were “nearly zero” articles, books or essays that explicitly dealt with the subject.</p>
<p>In his book <em>Worship of Creativity</em>, Samuel Franklin explores how creativity became an undeniable value and why technology leaders embraced it so enthusiastically. He spoke about why we are so fascinated by creativity, how Silicon Valley became its epicenter and how AI will reshape our relationship with creativity.</p>
<p><a href="#">Read the full interview.</a></p>We still have good things
<p><em>A place for comfort, fun and distraction to brighten your day. Have an idea? </em><a href="mailto:[email protected]"><em>Please contact us</em></a><em>.</em></p>
<ul>
<li>It took five days and 19,000 marbles to assemble <a href="https://www.youtube.com/watch?v=9bSxuIIozb8">this amazing marble run</a>.</li>
<li>A border collie transforms into <a href="https://www.youtube.com/watch?v=ISYbX5SpuX0">an entire zoo</a> with adorable animal masks.</li>
<li><a href="https://www.youtube.com/@grainydaysss">Grainy Days</a> is a YouTube channel featuring adventures captured beautifully on film photography.</li>
<li>Scientists created an <a href="https://a-hidden-infrastructure.spun.earth/story/mycorrhizal-infrastructure-map">interactive map of underground fungal networks</a> long enough to reach the sun a billion times.</li>
</ul>Source: www.technologyreview.com