OpenAI launches always-on dots agents & Moonshot AI strips cyber training data - Tech News (Oct 25, 2026)

Elon Musk has now said, in his own words, that he was the one who restricted Starlink service in Ukraine before it came back online. We'll get into why that admission matters. Welcome to The Automated Daily, tech news edition. The podcast created by generative AI. I'm TrendTeller, and today is October 25th, 2026. Today's lineup includes a new breed of AI agents from OpenAI, a security scramble affecting a big chunk of the web, and the widening tech divide between Washington and Beijing. Let's get started.

OpenAI launches always-on dots agents

We start with OpenAI, which used its developer event to unveil something it calls dots. These are always-on agents built to chase a goal you give them across several applications, without you having to keep nudging them along. Think less chatbot, more digital colleague that keeps working in the background. It's a clear play for the enterprise market, where rivals are racing to offer the same thing, and it marks another step away from AI that waits for instructions toward software that acts on its own.

Moonshot AI strips cyber training data

Staying with AI and its risks, an update on Moonshot AI. The Chinese startup says that for its new Kimi K3.1 Pro model, it deliberately pulled cyber-related material out of both the training data and the environments used to fine-tune the model's behavior. The company says the model still codes well in general, but is noticeably weaker at offensive hacking. It's an interesting trade-off: deliberately limiting a capability rather than just adding guardrails on top.

PHP vulnerabilities trigger emergency patching

On the security front, the PHP story continues. Teams around the world are now in emergency patch mode after the FBI warned the flaws could let attackers run code remotely on affected servers. With PHP powering roughly seven in ten websites, the cleanup itself is causing disruption, with some outages and reduced service as organizations push fixes out. Short-term pain, but much better than the alternative.

Ukraine strikes Russian tech data centers

Turning to the war in Ukraine, there's a new development involving digital infrastructure. Ukraine reportedly struck major data centers belonging to Yandex, Sberbank and VKontakte. Services were disrupted, and the attacks also hit Russian AI research and other digital operations. It's a stark reminder that server farms are now treated as strategic targets, not just civilian buildings full of computers.

Musk admits restricting Starlink in Ukraine

Also tied to Ukraine: Elon Musk said he personally was responsible for restricting Starlink connectivity there before service was restored. Starlink remains a lifeline for Ukrainian communications, so hearing that one private executive could switch it down has reopened an uncomfortable debate. How much power should a commercial satellite operator have over a country's military communications in wartime? There's no easy answer, but the question isn't going away.

US declares helium a critical material

Now to the economic tug of war between the United States and China. First, Washington has officially designated helium as a critical material. That might sound odd, but helium is essential for cooling and other specialized processes in chip manufacturing. The label could shape future policy on stockpiles and on securing supply from home or from allies.

Treasury proposes China tech investment limits

Second, the US Treasury has proposed rules to limit high-tech investment in China and other designated adversaries, focusing on technologies with national-security weight. If finalized, they'd push American and Chinese investment worlds even further apart.

US and China form rival AI blocs

And zooming out, the two superpowers are effectively forming rival AI camps. Many countries want access to both ecosystems, the computing power, the models, the money, without angering either side. That balancing act could ultimately decide which standards and infrastructure the rest of the world adopts. China, for its part, keeps building out regional capacity, with a large AI and cloud architecture gathering in Chengdu this week focused on combining supercomputing, AI computing and emerging fields like quantum and optical computing.

Meta's AI spending rattles markets

Over to the money side of AI. Following up on Meta's results, the numbers underlined just how expensive AI infrastructure has become, and how much of that build-out is now funded with borrowed money. That sparked a selloff in AI-linked stocks and renewed the core investor question: will the payoff ever justify the bill? The pressure spilled over onto other big AI players and tech valuations more broadly.

Barclays cuts jobs, boosts AI spending

Meanwhile, Barclays is the latest big name to trim its workforce while moving spending toward AI. It fits a pattern we're seeing across banking, tech and consumer companies: budgets shifting from people to processors. Whether that bet pays off is, again, the open question.

Tech firms consult philosophers on AI ethics

Finally, a more reflective story. Tech companies are increasingly sitting down with philosophers and religious leaders to talk about how AI systems should tell right from wrong. As worries grow about what powerful AI could do, executives are looking for moral frameworks beyond the engineering team. The tricky part is turning centuries of ethical and spiritual thinking into something a model can actually follow, and nobody has cracked that yet.

That's it for today's episode of The Automated Daily, tech news edition. From autonomous agents to helium shortages to the ethics of machines, it was a busy one. Thanks for listening, I'm TrendTeller, and I'll see you tomorrow with more.

OpenAI launches always-on dots agents & Moonshot AI strips cyber training data - Tech News (Oct 25, 2026)

Our Sponsors

Today's Tech News Topics