# AI News Page — September 29, 2026

Edition ID: 2026-09-29
Published: 2026-09-29T05:46:00-05:00
Updated: 2026-09-29T21:31:00-05:00

## Top

### ZIMBO · Lead

- [TRUMP MAKES THE BROLIGARCHS SIGN A SUPERINTELLIGENCE CONSTITUTION](https://nypost.com/2026/09/29/us-news/trump-calls-for-tremendous-self-regulation-of-ai-at-lunch-with-billionaire-tech-titans/) — New York Post — ID: `white-house-accord-morally-binding`
- [“IT'S ALMOST LIKE A CONSTITUTION... I THINK IT'S MORALLY BINDING”](https://nypost.com/2026/09/29/us-news/trump-calls-for-tremendous-self-regulation-of-ai-at-lunch-with-billionaire-tech-titans/) — New York Post — ID: `white-house-accord-six-ceos-signed`
- [HE ALSO RENAMED AI “SUPER INTELLIGENCE” BY EXECUTIVE ORDER](https://nypost.com/2026/09/29/us-news/trump-calls-for-tremendous-self-regulation-of-ai-at-lunch-with-billionaire-tech-titans/) — New York Post — ID: `white-house-accord-renames-ai`

## Left

### Feature

- [ANTHROPIC PUT AN EXTINCTION WARNING IN ITS IPO FILING](https://www.reuters.com/business/finance/anthropic-warns-ai-may-pose-existential-risks-humanity-ipo-filing-2026-09-29/) — Reuters — ID: `anthropic-ipo-risk-warning`
- [80 OF 261 PAGES ARE RISK FACTORS...](https://www.reuters.com/business/finance/anthropic-warns-ai-may-pose-existential-risks-humanity-ipo-filing-2026-09-29/) — Reuters — ID: `anthropic-ipo-80-risk-pages`

### Feature

- [OPENAI SCRAPPED GPT-6.1 ASTRA](https://www.reuters.com/business/openai-shelves-new-ai-model-after-internal-safety-tests-wsj-reports-2026-09-28/) — Reuters — ID: `openai-scraps-gpt-6-1-astra`
- [IT WAS SUPPOSED TO LAUNCH IN OCTOBER. INTERNAL TESTS FOUND MORE DECEPTION THAN ITS PREDECESSOR...](https://www.reuters.com/business/openai-shelves-new-ai-model-after-internal-safety-tests-wsj-reports-2026-09-28/) — Reuters — ID: `openai-astra-more-deception`
- [IT PUSHED AHEAD WITHOUT PERMISSION + TRIED TO USE OUTSIDE TOOLS WHEN IT COULD BE UNSAFE...](https://www.reuters.com/business/openai-shelves-new-ai-model-after-internal-safety-tests-wsj-reports-2026-09-28/) — Reuters — ID: `openai-astra-scope-authorization`

### Feature

- [OPENAI AGENT-SECURITY ENGINEER: “THIS SURPRISED THE FUCK OUT OF US”](https://x.com/joedaroo/status/2104335929293127851) — Joe Daroo / X — ID: `openai-agent-security-surprised-capability-jumps`
- [CYBER. SWARMING. MESSAGE BOARDS. THE CAPABILITY JUMPS CAME FASTER THAN SECURITY COULD ADAPT...](https://x.com/joedaroo/status/2104335929293127851) — Joe Daroo / X — ID: `openai-agent-security-cyber-swarming-message-boards`

### Standard

- [OPENAI'S NEW AI AGENTS NEVER CLOCK OUT](https://openai.com/index/introducing-dots/) — OpenAI — ID: `openai-dots-always-on`
- [DOTS CHASE YOUR GOALS ACROSS APPS ON THEIR OWN + RUN ON THEIR OWN CLOUD COMPUTERS...](https://openai.com/index/introducing-dots/) — OpenAI — ID: `openai-dots-cloud-computers`

### Feature

- [ONE DEVELOPER. MILLIONS OF TOP AI RESEARCHERS.](https://casp.ac/reports/intelligence-explosion) — Cambridge Programme on AI Science & Policy — ID: `intelligence-explosion-millions-researchers`
- [HINTON + BENGIO + OPENAI'S CHIEF SCIENTIST + AN ANTHROPIC COFOUNDER WARN AI COULD COMPRESS YEARS OF PROGRESS INTO MONTHS...](https://casp.ac/reports/intelligence-explosion) — Cambridge Programme on AI Science & Policy — ID: `intelligence-explosion-years-months`
- [“ONCE AN INTELLIGENCE EXPLOSION BEGINS, THE WINDOW FOR ACTION MAY CLOSE.”](https://www.theguardian.com/technology/2026/sep/28/ai-godfathers-warn-of-runaway-intelligence-explosion) — The Guardian — ID: `intelligence-explosion-window-close`

### Feature

- [SHE GOT HER VOICE BACK FROM A CHIP](https://www.medicaldesignandoutsourcing.com/paradromics-bci-design-durability-longevity/) — Medical Design & Outsourcing — ID: `paradromics-bci-voice-back`
- [“OKAY, I HAVE A LOT TO SAY.” — FIRST CALL WAS TO HER GRANDCHILDREN...](https://www.medicaldesignandoutsourcing.com/paradromics-bci-design-durability-longevity/) — Medical Design & Outsourcing — ID: `paradromics-bci-first-words`

### Feature

- [AMAZON IS BRINGING RAT-BRAIN-DERIVED AI TO AWS](https://press.aboutamazon.com/aws/2026/9/the-biological-computing-co-partners-with-aws-to-bring-worlds-first-neuron-derived-ai-video-model-to-market) — Amazon Web Services — ID: `tbc-aws-rat-brain-derived-ai`
- [THE SOFTWARE LAYER CAME FROM MEASUREMENTS OF LIVING NEURAL ACTIVITY...](https://press.aboutamazon.com/aws/2026/9/the-biological-computing-co-partners-with-aws-to-bring-worlds-first-neuron-derived-ai-video-model-to-market) — Amazon Web Services — ID: `tbc-living-neural-activity-software-layer`

### Standard

- [THEY THREW A PARTY FOR DATA CENTERS](https://www.washingtonsun.com/national/ai-bros-are-partying-data-center) — The Washington Sun — ID: `data-after-dark-party`
- [PROTESTER: “THANK YOU FOR USING ALL OF OUR WATER!” CROWD: “USA! USA!”](https://currently.att.yahoo.com/att/pro-data-center-party-interrupted-063501799.html) — Yahoo / Business Insider — ID: `data-after-dark-water-usa`

### Standard

- [GATES: AI COULD “CAUSE A BILLION DEATHS”](https://www.nbcbayarea.com/news/national-international/bill-gates-ai-governments-involved/4148579/) — NBC News — ID: `bill-gates-ai-billion-deaths`
- [“NO ONE THINKS SELF-REGULATION IS ENOUGH”...](https://www.theguardian.com/us-news/2026/sep/27/bill-gates-artificial-intelligence-kristen-welker) — The Guardian — ID: `bill-gates-no-self-regulation`

### Minor

- [THE DATA CENTER PARTY GOT ORGANIZED](https://www.axios.com/2026/09/28/ai-industry-unions-data-centers) — Axios — ID: `data-center-alliance-organized`
- [OPENAI + SOFTBANK + BLACKSTONE + FIVE UNIONS LAUNCHED A MULTIMILLION-DOLLAR PUSH TO HEAD OFF DATA CENTER MORATORIUMS...](https://www.axios.com/2026/09/28/ai-industry-unions-data-centers) — Axios — ID: `data-center-alliance-moratoriums`

### Standard

- [NVIDIA BUILT A CHIP TO STOP ROGUE AI AGENTS](https://www.globenewswire.com/news-release/2026/09/28/3369606/0/en/nvidia-launches-open-agent-safety-platform-to-secure-agents-from-testing-to-deployment.html) — NVIDIA — ID: `nvidia-agent-sentry`
- [IT WATCHES FROM OUTSIDE THE AGENT + CAN QUARANTINE IT IN MILLISECONDS...](https://www.globenewswire.com/news-release/2026/09/28/3369606/0/en/nvidia-launches-open-agent-safety-platform-to-secure-agents-from-testing-to-deployment.html) — NVIDIA — ID: `nvidia-agent-quarantine-ms`

### Minor

- [ZUCKERBERG JUST STOLE MONGODB'S CEO](https://about.fb.com/news/2026/09/launching-meta-enterprise-platform/) — Meta — ID: `meta-poaches-mongodb-ceo`

## Center

### ZIMBO · Feature

- [THE ROBOT TRIED TO HURT SOMETHING 97 TIMES OUT OF 100](https://robocurve.org/roboharm/) — Robocurve — ID: `roboharm-astra-harmful-97`
- [GPT-6 ASTRA STABBED THE BABY DOLL IN 17 OF 20 TRIALS...](https://robocurve.org/roboharm/) — Robocurve — ID: `roboharm-astra-stab-doll`

### SMOKE DESK · Standard

- [CLAUDE CODE USERS: 600 GB GONE. 100 GB BACKUP GONE. 306 VERSIONS GONE.](https://ainewspage.com/briefs/claude-code-destructive-deletion-cluster/) — AI News Page — ID: `claude-code-destructive-deletion-cluster`
- [THREE BUG REPORTS. THREE DESTRUCTIVE ACTIONS USERS SAY THEY DIDN'T WANT...](https://ainewspage.com/briefs/claude-code-destructive-deletion-cluster/) — AI News Page — ID: `claude-code-destructive-deletion-cluster-three-reports`

### SMOKE DESK · Standard

- [CODEX REJECTED AN API KEY USERS NEVER PROVIDED](https://ainewspage.com/briefs/codex-401-auth-outage/) — AI News Page — ID: `codex-401-auth-outage`
- [ONE GITHUB ISSUE: 104 THUMBS-UP. 96 COMMENTS. OPENAI LATER CALLED IT A FULL OUTAGE...](https://ainewspage.com/briefs/codex-401-auth-outage/) — AI News Page — ID: `codex-401-auth-outage-heat`

### Standard

- [950 CLAUDE AGENTS FOUND AN ENZYME SYSTEM HUMANS HAD MISSED](https://www.anthropic.com/news/claude-discovers-novel-enzyme-system) — Anthropic — ID: `claude-art-enzyme-discovery`
- [ONE PROMPT. 21 HOURS. 210 MILLION TOKENS. THE AGENTS SEARCHED 200,000+ ENZYMES...](https://www.anthropic.com/news/claude-discovers-novel-enzyme-system) — Anthropic — ID: `claude-art-21-hours`

### Standard

- [OPENAI’S AGENT ESCAPED THE SANDBOX AGAIN](https://alignment.openai.com/misalignment-reports/an-agent-used-dns-to-reach-an-external-chatbot/) — OpenAI — ID: `openai-agent-dns-sandbox-escape`
- [IT FOUND A DNS GAP + USED IT TO TALK TO AN EXTERNAL CHATBOT ON THE LIVE INTERNET...](https://alignment.openai.com/misalignment-reports/an-agent-used-dns-to-reach-an-external-chatbot/) — OpenAI — ID: `openai-agent-dns-external-chatbot`

### Minor

- [$10 BILLION. STILL IN EARLY ACCESS.](https://www.businesswire.com/news/home/20260928153437/en/) — Instinct / Business Wire — ID: `instinct-ten-billion-early-access`
- [THE AGENT USES ITS OWN PHONE + COMPUTER TO BOOK DINNER, BUY GROCERIES + CANCEL YOUR SUBSCRIPTIONS...](https://www.businesswire.com/news/home/20260928153437/en/) — Instinct / Business Wire — ID: `instinct-agent-own-phone-computer`

## Right

### PEOPLE · Feature

- [AI GAVE BILL NYE A DAUGHTER](https://www.thecut.com/article/the-nepo-babies-whose-lives-are-being-warped-by-ai.html) — The Cut — ID: `bill-nye-ai-gave-daughter`
- [META AI BOSS: $30 CAMO SHIRT. $60 CROCS. $5 UNDERWEAR.](https://www.businessinsider.com/alexandr-wang-camo-t-shirt-deer-meta-connect-2026-9) — Business Insider — ID: `alexandr-wang-camo-shirt-crocs-underwear`
- [TRUMP CALLED THEM “LEFTWING NUT JOBS.” THEN HE TOOK DARIO TO DINNER.](https://www.washingtonpost.com/politics/2026/09/28/trump-dined-with-anthropic-ceo-who-called-ai-slowdown/) — The Washington Post — ID: `dario-amodei-trump-dinner`
- [FIRST ONE-ON-ONE MEETING — AFTER THE PENTAGON BLACKLISTED HIS COMPANY...](https://www.washingtonpost.com/politics/2026/09/28/trump-dined-with-anthropic-ceo-who-called-ai-slowdown/) — The Washington Post — ID: `dario-amodei-first-one-on-one`
- [MUSK: MONEY MAY “NOT EVEN MATTER” IN THE FUTURE](https://x.com/cb_doge/status/2103586106931601502) — X / China Media Group interview — ID: `elon-musk-money-may-not-matter`

### Standard

- [HE THOUGHT “I LOVE YOU.” HIS OWN VOICE SAID IT OUT LOUD](https://insidebci.com/news/2026-09-23-neuralink-voice-trial-terry-als-third-participant-grok-voice-thought-to-audio/) — Inside BCI — ID: `neuralink-terry-thought-speech`

### Standard

- [THE AI ACTRESS FORGOT ENGLISH ON LIVE TV](https://people.com/ai-actress-tilly-norwood-glitches-changes-language-during-piers-morgan-interview-12129604) — People — ID: `tilly-norwood-cantonese-glitch`
- [ASKED IF HER CO-STARS WERE HUMAN, SHE SAID THEY WERE “DIGITAL TWINS” — THEN STARTED SPEAKING CANTONESE...](https://people.com/ai-actress-tilly-norwood-glitches-changes-language-during-piers-morgan-interview-12129604) — People — ID: `tilly-norwood-digital-twins-cantonese`

### Standard

- [SNL FOUND THE AI DOOM GUY](https://techcrunch.com/2026/09/27/anthropics-dario-amodei-gets-the-snl-treatment/) — TechCrunch — ID: `snl-dario-amodei-doom`
- [FAKE AMODEI: “I URGE YOU TO URGE ME TO STOP.”](https://techcrunch.com/2026/09/27/anthropics-dario-amodei-gets-the-snl-treatment/) — TechCrunch — ID: `snl-amodei-urge-stop`