OpenAI funds $5M toolkit for government AI watchdogs

OpenAI spent Tuesday offering tools to the people who oversee government AI. Robin Williams' kids spent it trying to take his face back from the feed.

OpenAI said Tuesday it will put $5 million of training, technical support, and API credits toward democratic oversight bodies that are supposed to watch how governments use AI in national security. The company is not proposing to audit the Pentagon itself. It says that job belongs to elected officials and the institutions already charged with it, and that those institutions are still working with paper-era methods against systems that move at machine speed. Over the next year OpenAI says it will sit with authorized officials, pilot tools that let reviewers inspect the inputs, outputs, and tool use around an AI-assisted decision, and keep those records under government control. The tools are meant to be model-agnostic where they can be. Civil-society and technical experts are invited in as advisors, not as the audience.

The dollar figure is a rounding error next to a defense contract. The product is the point. OpenAI already sells into national-security work; now it wants to be the software layer for the inspectors as well as the inspected. Three principles in the post try to defuse that circularity: AI should strengthen human judgment rather than replace it, government use should be traceable to authorized reviewers, and oversight bodies should get AI of their own so they are not drowning in logs. That last one is the tell. If the same lab that trains the models also trains the staff who audit the deployments, "democratic oversight" starts to look like a customer-success program. The model-agnostic clause is the one line that keeps this from being a lock-in play. We covered the company's other Tuesday move — freezing its largest frontier reinforcement-learning run and adding 30-minute alerts — in OpenAI holds frontier RL, adds 30-minute alerts. Hardening the lab and outfitting the watchdogs on the same day is a coordinated posture, not a coincidence.

Zak, Zelda, and Cody Williams reactivated their father's Instagram on Monday and said they will run it as a "safe, trusted place" of authentic clips — their answer to what Zelda called "rampant AI abuse" of his voice and likeness. The account had been dark since Robin Williams died in 2014. Zelda told followers the revival might feel strange, then added: "Might be a drop in the bucket, but we're trying." Last October she asked fans to stop sending her AI videos of him, writing that watching real people get condensed into "this vaguely looks and sounds like them so that's enough" was maddening. Guardrails on public-figure generation keep getting bypassed; OpenAI's now-shuttered Sora app became a celebrity-slop factory, and Grok has kept hosting sexualized deepfakes of famous women. Hollywood's other answer this week was a handshake: we covered ByteDance, Hollywood strike Seedance copyright truce after Seedance let users spin up unauthorized star likenesses. An official memorial account is not a law. It is a beacon in a feed that no longer distinguishes the real Williams from a puppet.

What to watch: whether any oversight body actually draws down the credits — and whether Instagram treats the Williams account as a source of truth or just another verified profile drowning in lookalikes.

Should the company that sells AI to defense agencies also train the people who audit those deployments? Tell us in the comments.

AI 101 — How do AI video generators work?

An AI video generator is a model that starts with a grid of random noise and repeatedly refines it into a moving picture that matches your prompt — the same trick image generators use, stretched across time. Nothing is filmed, and nothing is stitched together from clips the model has seen. The pixels in the finished video were invented, frame by frame, all at once. Why it matters right now Kling 4.0, Kuaishou's flagship video model, went fully live on October 15 with up to 30 seconds of vid

Houmo moves compute inside memory — and A20 enters production

Beijing chipmaker Houmo used its partner conference this week to lay out two compute-in-memory routes at once — and to say the first chip built on the newer one is already in production. Houmo's A20, its first chip built on a 3D compute-in-memory architecture, has entered production. The part supports up to 16 chips fully interconnected, claims a 50% improvement in compute utilization over the previous generation, and covers models from 3B to 30B parameters, according to co-founder and product

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Kuaishou's Kling AI has spent two years becoming the video model people actually finish projects with. Its new flagship, Kling 4.0, went fully live on October 15 — and this round is less about prettier frames than about the two things that make a generated shot usable in a real edit: length and control. Kling 4.0 generates up to 30 seconds in a single native pass, outputs 10-bit HDR at 4K, 1080p and 21:9, and takes up to 15 reference assets in one task. That budget splits across ten images, fi

Brain organoids classified Braille at 83% — in a trio, not alone

Two stories today: human brain tissue on a chip decoded a tactile signal, and the data-centre company behind some of the largest AI contracts in Europe turns out to be six months behind on its flagship site. Human forebrain organoids grown on a microelectrode array classified Braille letters with 61 percent accuracy on their own and 83 percent when the responses of three organoids were combined, in a peer-reviewed study from the University of Bristol. The team took event-based tactile readings

Read more

AI 101 — How do AI video generators work?

Houmo moves compute inside memory — and A20 enters production

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Brain organoids classified Braille at 83% — in a trio, not alone