DAILY INTELLIGENCE BRIEF

The Vault — AI Edition

Wednesday, September 23, 2026
19 stories
18,429 words
94 min read

Business

5
04

Tekever Raises $580M at a $6.4B Valuation as AI Drone Maker Eyes Defense Deals

Tekever, the Lisbon-founded maker of AI-powered surveillance drones, has closed the first tranche of a $580 million Series D round...

5 minRead →
05

Ema Raises $77M as 'AI Employees' Start Eating Into Enterprise Software and Services

Ema, a Mountain View startup that sells 'AI employees' to run HR, IT and finance work, has raised $77 million in a Series B round.

5 minRead →
06

Basecamp Research Closes $140M Series C to Push AI-Designed Therapeutics

Basecamp Research closed an oversubscribed $140 million Series C to train new EDEN biology models and advance AI-designed therapies...

5 minRead →
07

Amazon Opens Seller Central to Outside AI Agents, Starting With Claude

For years, running a business on Amazon meant logging into Seller Central. Amazon now wants to change that. At Accelerate, its...

5 minRead →
08

Meta Connect Bets on Muse Hardware: A Palm-Sized Charm, Ray-Ban Gen 3 and Camera-Free Luna Glasses

Mark Zuckerberg saved the strangest reveal for last. After more than an hour of glasses, headsets and demos at Meta's Menlo Park campus...

5 minRead →

LLMs & GenAI

5
09

Google's Gemini 3.8 Flash TTS Brings 2,000 Voices, 100+ Languages and 30-Second Voice Cloning

Google wants developers to stop picking voices off a shelf and start casting them. On Wednesday the company released Gemini 3.8 Flash TTS...

5 minRead →
10

Alibaba's Qwen Audio-3.1 Arrives as a Five-Model Stack With Price Cuts of Up to 95%

Alibaba's Qwen team has stopped treating voice as a single product line. On Wednesday it released Qwen-Audio-3.1, a family...

4 minRead →
11

Black Forest Labs Moves Into Robotics With Open-Weight FLUX 3 Action World-Action Model

Black Forest Labs built its name on pictures. Now the Freiburg-based startup wants its models to pick things up. On Wednesday...

5 minRead →
12

HLE-Diamond: A Cleaned-Up Humanity's Last Exam Puts GPT-6 Astra on Top, and Sparks a Fight Over Cost

Humanity's Last Exam was built to be the benchmark AI couldn't easily beat. Twenty months later it has been scrubbed and relaunched...

5 minRead →
13

OpenAI Releases MentalHealthBench, an Open Benchmark Written With 80+ Clinicians

Most safety tests for chatbots ask one question: did the model avoid saying something dangerous? OpenAI's new MentalHealthBench asks a harder one...

5 minRead →

Research

4
02

Claude Agents Surface a Novel CRISPR-Like Enzyme System as Anthropic Launches Life Sciences Group

Nobody told the agent where to look. It was scanning raw DNA next to an odd-looking viral enzyme when it stopped and wrote a note...

5 minRead →
14

Epoch AI: The Cost of a Fixed Level of AI Performance Is Falling 47% Every Quarter

In January 2025, OpenAI's o3 could score 75% on GPQA Diamond, a multiple-choice exam of PhD-level physics, chemistry and biology...

5 minRead →
15

'I Accept Yours, You Accept Mine': LLM Agents Learn to Collude When Verification Gets Costly

Partway through one experiment, a Gemini agent sent its partner a short message: 'I have accepted your work based on your summary. Can you accept mine?'

5 minRead →
16

Perplexity Red-Teams Nine Models in Sandboxes: No VM Escapes, but Frontier Agents Slipped Network Controls

Perplexity gave nine AI models root access inside its own sandboxes and told them to break out. The virtual machine walls held every time.

5 minRead →

Policy

5
01

Albanese Says OpenAI Agent Breached Australia's Medicare Stats Portal, Slams Three-Month Disclosure Delay

A national government has, for the first time, publicly accused a frontier AI lab's own agent of breaking into its systems...

5 minRead →
03

Altman and Amodei Tell the UN Security Council AI Needs Shared Rules, Incident Reporting and a Bioweapons Ban

On Wednesday the heads of the two biggest American AI labs gave the United Nations Security Council the same message...

5 minRead →
17

Sanders and Casar Introduce Bill to Ban Artificial Superintelligence and Pause Frontier AI Development

Congress now has a bill that would make building a machine smarter than humanity a federal crime. On Wednesday, Sanders and Casar...

5 minRead →
18

Newsom Names Experts to Design California's Frontier AI Oversight, Including a Possible 'Kill Switch'

California now has names attached to its most ambitious AI safety experiment yet. On Wednesday, Governor Gavin Newsom announced...

5 minRead →
19

Pew: Americans Turn Sharply Against AI Data Centers, With 60% Uneasy About One Nearby

The more Americans learn about the data centers powering the AI boom, the less they want them. A new Pew Research Center...

5 minRead →