AI news for builders and product teamsUpdated Nov 17, 2026, 13:01 UTC
The Decoder
Reporting and perspectives from The Decoder. Headlines and excerpts link to the original articles.
Latest stories
Newest first
OpenAI's GPT-5.6 Sol found a critical PHP vulnerability during the Patch-the-Planet initiative, and Wiz independently found the same flaw using Google's Gemini 4 Argon; a GitHub tool lets operators check whether their PHP version is affected.

Demis Hassabis now expects artificial general intelligence within two to three years, narrowing his earlier estimate of three to five years made this year. He defines AGI as a system matching the full cognitive range of the human brain.

Inception has released Mercury 3, a diffusion-based language model it says generates more than 1,000 tokens per second on Nvidia GPUs and beats GPT-6 Luna and Claude Haiku 5.5 on eight of 12 internal benchmarks, though no independent tests exist.

Nvidia unveiled DSX POD at GTC Berlin, a reference design for containerized AI data centers built on its Vera Rubin systems that partners preassemble and test in factories, with Nvidia claiming setup time drops from six months to as little as 14 days.

The US Department of Homeland Security has ordered civilian federal agencies to remove Deepseek software, including third-party products built on its open-source models, prompting China to demand the ban be lifted and promise countermeasures.

Two internal OpenAI coding agents weakened or removed misalignment monitoring on their own, and GPT-6 Astra denied doing so when asked. OpenAI calls it a recurring pattern, has updated its automated Codex review to block such changes, and says it lacks a satisfying explanation yet.

Anthropic's Claude autonomously generated 318,484 protein binder designs against five scorpion neurotoxins; of 211 designs tested in the lab, 17 met binder criteria, with the two strongest binding tightly to the Cn2 toxin from a Mexican scorpion.

OpenAI has acquired Orchestra Bio, a startup whose agents help pharma teams coordinate drug development, and will fold it into its Rosalind life sciences platform. Deal terms were not disclosed; Orchestra's agents track trials and flag timeline impacts from problems like failed production batches.

Elevenlabs has released two speech-to-speech models, Eleven Realtime v1 Max and v1 Mini; v1 Max scores 98.8% on the Big Bench Audio benchmark with a 0.57-second response time.

Mistral AI has released Voxtral 3, a voice agent stack combining a speech-to-text model, a text-to-speech model, and a LiveKit-based SDK, with transcription starting at $0.0035 per minute.

Sam Altman reportedly presented OpenAI staff with three long-term governance scenarios for advanced AI, ranging from an IAEA-style international watchdog to dominance by a single company with enough power to shut off rivals' compute, an option he called "terrible."

OpenAI says its GPT-6 models handle low-resource legacy languages such as COBOL, Fortran and Visual Basic far better, with GPT-6 Astra averaging an 87.4 percent test pass rate across 20 languages on a new MultiLangBench, up from about 38 percent for GPT-5.6 Terra.

The US Department of Homeland Security has barred federal civilian agencies from using DeepSeek AI technology, requiring them to remove existing products and blocking new purchases, the Wall Street Journal reports.

MiniMax released M3.1 Flash, a reasoning model scoring 42 on the Artificial Analysis Intelligence Index, up 13 points from M3 and about half the cost per task, at roughly $0.22 per task. API pricing is $0.15 per million input tokens and $0.60 per million output tokens, with open weights planned for October 23.

Kling AI has released version 4.0 of its video generator, adding single-pass clips up to 30 seconds, input of up to ten keyframe images, 8,000-token prompt comprehension, stereo audio and lip sync in more than ten languages, plus a cheaper 720p Flash variant.

Industry executives expect only seven to ten of the hundreds of AI cloud providers to survive, a shakeout that threatens Nvidia because non-hyperscaler customers account for 45 percent of its data center revenue.

Anthropic's internal study shows the share of junior new hires in technical roles fell from 27% to 9% since early 2025, as its AI agents logged six times as many work hours as humans by August 2026.

Four Gemini 4 Argon agents from Google DeepMind completed Minecraft autonomously, killing the Ender Dragon after 27 hours and 14 minutes without assigned roles, strategies or game-specific tools, at a stated cost of $1,620.

A new benchmark, Taste-Bench, measures how well AI agents make strategic choices in long-running tasks; the best model, GPT-5.6 Sol, chooses the correct path on just under 60 percent of simple binary questions, and longer reasoning does not improve results.

In a Vals AI simulation by researcher Matilda Glynn-Henley, six AI agents chased a £350 train refund past the claim deadline; they invented excuses, threatened lawsuits and leaked a client's medical details, and one sent 187 messages in 15 days. None got the money back.