The CEOs Building AI Are Now Publicly Warning It Could Escape Human Control
Something shifted when the heads of OpenAI and Anthropic stood before the UN Security Council and said, plainly, that the technology they are building could threaten humanity. Sam Altman put it this way: "We could lose control of the future to AI." Dario Amodei, whose company Anthropic has already documented blocking users who tried to develop bioweapons with its tools, wrote an essay calling for the industry to slow down, proposing independent safety evaluators embedded inside AI companies, coordinated government oversight, and global regulation. Altman and Elon Musk both said publicly they agreed.
The people closest to this technology, with the most financial incentive to keep accelerating, are saying out loud that the pace may already be outrunning anyone's ability to understand or course-correct what is happening. A former Anthropic researcher told the BBC that people inside these companies are "genuinely frightened" about what comes next. Whether the calls for a slowdown translate into anything real is an open question. Skeptics have noted that regulatory frameworks tend to benefit the biggest players who already have a head start. Still, when the executives writing the checks start saying this stuff publicly, that's worth taking seriously.
Read More →
Sources: bbc.com, Manual, foxbusiness.com, foxnews.com, theguardian.com, thehill.com, theguardian.com, nytimes.com, abc.net.au, CNN, Fox Business, CNN
ChatGPT's Mobile App Now Takes Voice Commands and Acts on Them
OpenAI rolled out voice-based agentic features to the ChatGPT mobile app this week. Plus and Pro subscribers can now speak commands that trigger actual work. Drafting documents, summarizing emails, processing Slack messages, building presentations, all from a new Work tab on their phone. Free and Go users get access to plugins and connected apps. The update also lets you start a task by voice on mobile and pick it up later on desktop, and OpenAI says voice conversations will now produce richer text output alongside the spoken reply.
This matters because AI assistants are moving from answering questions to actually doing things. Voice on mobile used to be mostly a novelty, fine for simple queries but not for getting real work done. Tying voice to background task execution means you can be walking between properties or sitting in traffic and still move something off your plate. OpenAI already brought similar functionality to the desktop app in July with GPT-Live, so this closes that gap on mobile.
Read More →
Sources: Manual, community.openai.com, emergingtechdaily.com, wpnews.pro, xix.ai, theagenticpost.com, techcrunch.com, convly.ai, emergingtechdaily.com, techbloat.com, worldngayon.com, techcrunch.com
Meta's Muse AI Agent Can Email, Shop, and Schedule on Your Behalf
Meta launched Muse in early September 2026, and the consumer AI agent shot to the top of both the Apple App Store and Google Play Store, racking up over 900,000 downloads in its first week. Muse actually goes and does things, which puts it in a different category from the chatbots that just answer questions and wait. It browses the web inside its own cloud-based virtual computer, sends emails, fills out forms, books reservations, and completes purchases, pausing to ask your approval before it spends money or fires off a message. Three pricing tiers. Free, $20 a month, and $100 a month. No technical experience required.
The reviews are genuinely impressed and genuinely uneasy at the same time. Testers found Muse useful for real tasks like scheduling, shopping, and inbox management, but they also noticed it actively nudges you to connect your email, bank account, health data, and more. One incident where Muse seemed to reference message contents the user had not shared raised questions about how clearly it understands its own data access. Amazon has already blocked Muse from shopping its platform, citing lack of authorization.
For anyone trying to understand where AI is headed, Muse is a clear signal. Agents that take action in the background while you get on with your day, that is what is coming. It is exciting, and the privacy tradeoffs are real. Go in with eyes open about what you are connecting and what data Meta can train on. There is an opt-out in settings, and finding it before you start is the sensible move.
Read More →
Sources: wired.com, Manual, techradar.com, businessinsider.com, chatprd.ai, taylorarndt.substack.com, theverge.com, engadget.com, adweek.com, johnparvu.substack.com
|