Lead story / launch

OpenAI adds editable GIFs to ChatGPT, with control over individual frames

OpenAI released Images 2.6 on November 4, adding animated GIF creation to ChatGPT from photos or text prompts. Users can inspect an animation frame by frame, change playback speed and revise a single frame before sharing. OpenAI says the release is available globally in ChatGPT and Codex across all subscription tiers, on desktop, mobile and web.

The animation works like a flipbook: generate related still images, then string them together. In a dancing-cat demonstration described by The Decoder, the character stayed recognizable enough for the sequence to work. That is an observation about the demo, not a measured result across other characters or scenes. OpenAI pitches the feature for animated mascots, social reactions, product loops and storyboards.

Developers get two API models: GPT-Image-2.6 Flare, positioned for speed, and GPT-Image-2.6 Sunburst, positioned for greater detail and consistency through edits at medium or high quality. OpenAI says Images 2.6 generates images 50% faster than its predecessor. Those performance claims remain company claims; neither model appeared on Artificial Analysis’s image-generation or editing leaderboards in OrcaRouter’s November 5 check.

The API also adds region markup for targeted edits and SVG export for scalable designs. A separate complexity setting controls scene-planning effort rather than visual detail, so developers can choose image quality independently. OpenAI recommends low complexity for most editing tasks and medium or high complexity for scenes with many elements or constraints.

policy

AI critics win congressional seats after pushing for tighter controls

Abdul El-Sayed and Manny Rutinel won seats in Congress after campaigning for tighter AI controls. WIRED also reports that voters in nine states rejected data-center development plans, approved moratoriums or sought more community input. Incoming governors in Georgia and Texas backed construction pauses. Those wins strengthen critics’ political voice, but campaign positions are not restrictions already in force, and no new federal AI policy was announced.

research

Anthropic uses Claude to map a shell built from 96 copies of a viral protein

Anthropic says Claude helped researchers resolve the first complete vault structure from prokaryotic life, using a protein from a bacteria-infecting virus. Its optimized preparation took about 90 minutes, versus roughly three days for the conventional method it described. Researchers guided the work, and a Claude-written protocol error caused a two-week delay. The shell’s biological purpose remains unknown; the findings were submitted as a preprint.

security risk

OpenAI says its AI pipeline built 948 Linux root-access exploits

OpenAI says its prototype pipeline generated 948 exploits that let an existing low-privilege Linux user gain root access, plus 291 information-leak demonstrations. It tested stock desktop virtual machines, not systems weakened for testing. The run consumed $1,213,430 in API credits, not an all-in research cost. OpenAI has begun reporting critical flaws; candidate patches still need maintainer review.

Also worth knowing

•  Musk endorses space data centers after Democratic election gains

•  Stanford researchers classify 44% of tested AI findings as “expired”

•  MIT researchers build robot memory that checks uncertain descriptions

•  Audit finds a target mismatch in OpenAI’s 45× Linux bounty estimate

•  Uber says AI agents booked just four rides during Halloween week

•  Democrats gain new leverage over AI with control of Congress

•  UT Arlington study finds safety tools miss most targeted scam requests

•  OpenAI announces Pierre Computer acquisition to add code storage to Codex

Daily tool drop

6 AI tools worth knowing today

The Internet Had a Point

Meme from the stashAI, Meet the Korean Historian

Reader check-in

Help shape tomorrow's briefing

One click tells us what to keep, improve, or tighten.