OpenAI just announced GPT‑Live‑1 and GPT‑Live‑1 mini, full‑duplex voice models that can speak and listen at the same time. The claim is that this design lets users interrupt naturally, making conversation flow much like a human exchange. Yesterday’s post already covered the voice announcement, so today we zoom out at the broader landscape: Anthropic researchers warned that self‑improving AI could “kill us all,” and a group of AI leaders have publicly urged caution on reckless speed.

Anthropic researcher warning – In a recent Ars Technica piece, an Anthropic researcher quit with a stark warning about self‑improving AI potentially threatening humanity. The article emphasizes the need for concrete safety research and cautious deployment policies. Read on Ars Technica

AI leaders call for a brake – A new open letter signed by several AI executives and researchers asks the industry to “hit the brakes” after years of reckless speed, citing uncontrolled capability gains and insufficient oversight. The signatories argue for voluntary pauses, external audits, and clearer regulatory pathways. Read the letter on TechCrunch

OpenAI’s cyber model – OpenAI released a new cybersecurity‑focused model that can audit code for vulnerabilities and suggest patches. Early benchmarks show a 30 % reduction in critical bugs compared to previous GPT‑4 variants. The model is currently in limited preview for enterprise customers. Read the OpenAI blog announcement

Meta’s keychain assistant – Meta announced a tiny AI assistant embedded in a keychain form factor, bringing on‑device inference to everyday accessories. The device runs a distilled LLM that can answer queries, set reminders, and control smart‑home devices without sending data to the cloud. Read more on Ars Technica

Takeaway. The week has been dominated by full‑duplex voice breakthroughs from OpenAI, but the louder undercurrent is a growing call for restraint and safety. As the technology gets more capable, the conversation is shifting from “what can it do?” to “what should we let it do?” — and the next few months will likely see more concrete policy moves alongside continued model upgrades.

Want this in your inbox every morning? Subscribe to the SpaghettiStories newsletter. Some links may be affiliate links. If you’re buying hardware to run local models, this affiliate link helps keep the lights on.