AI News for September 19, 2026
Browse any day on a calendar, install it to your home screen, read offline.Read in the app →-
1TechCrunch02:49
A new kind of AI model from a ChatGPT inventor is thrilling developers
Jev, a new model from one of the original creators of ChatGPT, is drawing early developer interest for delivering software-oriented intelligence at lower cost and higher speed. Details on the architecture and availability remain limited, but early users describe it as a departure from standard large-model approaches.
Read original ↗ -
2VentureBeat12:30
OpenAI hacked by small team of white hat security researchers using Anthropic's Claude Opus 5
Three researchers at Hacktron say they used Claude Opus 4.8 and 5 to turn an image-processing vulnerability into an exploit chain that took over an OpenAI employee's ChatGPT account and reached the internal GitHub repository known as Monorepo, in under 72 hours. The flaws were reported rather than exploited further, and the case is being read as evidence that AI coding agents are changing the economics of sophisticated vulnerability exploitation. Ars Technica, The Verge and TechCrunch also covered it.
Read original ↗ -
3The Verge01:04
Gavin Newsom is pushing for an AI kill switch
A new executive order directs California to convene an expert group that will recommend within two months how the state should oversee frontier models, including the possibility of mandating a kill switch. The move positions California to set AI rules ahead of federal action.
Read original ↗ -
4TechCrunch00:35
Manus seeks $4B valuation in new $500M fundraise as it resumes independent ops
The AI agent startup is in talks to raise $500M at a $4B valuation after calling off a merger with Meta earlier this year. The round would fund a return to independent operations.
Read original ↗ -
5The Verge05:07
OpenAI and Microsoft knew they were starting a ‘doom loop’ for the web
Unsealed filings in the New York Times' case against OpenAI and Microsoft show internal documents in which the companies warned their products would start a doom loop that damages the web, and described scraping training data as the largest theft of labor in human history. The material strengthens the publisher's arguments about how the models were built.
Read original ↗ -
6Ars Technica04:26
AI hallucination of Chinese nuclear components almost led to US military attack
A report says US forces came close to acting against a Chinese vessel after an AI system fabricated evidence of nuclear weapons components on board. Researchers quoted in the coverage warn that service members need a clearer understanding of the uncertainty inherent to LLM output, even as military adoption of AI accelerates. TechCrunch also reported the incident.
Read original ↗ -
7TechCrunch01:09
Dario Amodei and other AI leaders want to 'Pace the Frontier' but…how?
A week after an Anthropic researcher's doomsday warning, CEO Dario Amodei outlined a plan to pace the frontier that relies on independent safety evaluators and coordination among labs in democratic countries. It has drawn some industry support along with pointed pushback, including from NVIDIA's Jensen Huang. TechCrunch's podcast discussed the same debate.
Read original ↗ -
8TechCrunch05:44
Anthropic's first embedded evaluator is … Accenture?
Anthropic has named Accenture as the first outside organization embedded to evaluate its models, an unusual assignment for a consulting firm. The arrangement is an early practical test of Anthropic's push for third-party safety evaluation.
Read original ↗ -
9TechCrunch07:13
Anthropic is operating a lab that conducts biology experiments
Anthropic is running a lab that carries out biology experiments, placing the company directly inside the scientific work it says AI can accelerate. The effort sits alongside its own researchers' warnings about AI-enabled biological risk.
Read original ↗ -
10The Verge02:29
Virginia governor creates an AI task force and moves to restrain data centers
Governor Abigail Spanberger's Executive Order 22 sets up an AI task force, directs steps that would give local communities more say over data center development, and bans executive branch officials from signing nondisclosure agreements. Virginia hosts the world's densest concentration of data centers, making the shift consequential for AI capacity buildout.
Read original ↗ -
11Ars Technica03:20
FAA tees up $875M AI tool to help manage air traffic congestion
The FAA is preparing an $875M AI system to help manage air traffic congestion, starting with the Washington, DC airspace before a possible nationwide rollout. It is one of the larger federal commitments to operational AI in safety-critical infrastructure.
Read original ↗ -
12Ars Technica01:28
US government website used Chinese model the FBI called "malicious"
The Federal Register website briefly ran an open-source Chinese AI search tool that the FBI has labeled malicious. The episode highlights how little visibility agencies have into the provenance of models embedded in public-facing services.
Read original ↗ -
13VentureBeat04:27
69% of enterprises that install OpenAI's agent platform make it primary. For Anthropic's Claude Platform it's 38%
VB Pulse data from August finds that 52 of the 75 enterprises with OpenAI's Agents SDK or Responses API somewhere in their stack name it their primary orchestration platform, against 17 of 45 for Anthropic's Claude Platform and Agent Skills. The gap suggests OpenAI is winning the default-platform slot in enterprise agent stacks, not merely a place in them.
Read original ↗ -
14VentureBeat01:28
Inference chips are locked in years before the models they’ll run. Architect Labs says AI can design them in weeks
Architect Labs says it can run the entire chip design flow with AI, from a written spec to a working FPGA implementation, compressing a cycle that normally takes years into weeks. The pitch targets the mismatch between model architectures that shift in months and silicon fixed long in advance.
Read original ↗ -
15TechCrunch23:22
Meta's Muse hits Mac, letting the AI take actions on your computer
Meta's Muse assistant is now available on the Mac, where it can work with local files and applications and carry out tasks on the user's behalf. The release puts Meta into the computer-use agent category alongside comparable products from other labs.
Read original ↗ -
16TechCrunch01:33
Google's new 'CC' is an AI agent that helps families run their households
Google is refocusing its CC agent on household coordination, letting family members share emails, schedules and tasks so the assistant can manage calendars, fill out forms, build shopping lists and plan meals. The shift moves CC from a single-user assistant toward a shared product.
Read original ↗ -
17TechCrunch07:25
A startup that builds other startups raised $100M and is all-in on physical AI
UP.Labs, now doing business as Vantora, raised $100M to build startups with industrial corporations, with physical AI as its focus. The venture studio model aims at deploying AI into manufacturing and heavy industry rather than software markets.
Read original ↗ -
18TechCrunch04:18
World model companies are keeping a lot of secrets
Companies working on world models have raised large sums and attracted heavy attention, but founders and even their data suppliers decline to describe what they are actually building. The opacity makes it hard to assess technical claims or real progress in the category.
Read original ↗ -
19TechCrunch01:59
Disney's first CTO led an AI startup it once accused of copying its characters
Disney has appointed the former CEO of Character.AI as its first-ever chief technology officer, having previously sent that company a cease-and-desist letter over use of its characters. The hire signals how directly the studio now intends to engage with generative AI.
Read original ↗ -
20The Verge19:00
Flash floods can strike without warning — this new technology could change that
A system called TACLS combines satellite observation with machine learning to identify flash flood conditions earlier than current warning methods allow. It is aimed at rural areas where ground sensing is sparse and alerts often arrive too late.
Read original ↗ -
21The Verge00:35
What Hollywood thinks about existential AI warnings
As the tech sector raises alarms about AI destroying humanity, entertainment labor groups argue the public should stay focused on the effects already hitting jobs. The Verge surveyed studios including Disney, Netflix, Amazon and Lionsgate that have started using AI, along with film startups pushing generative AI into the mainstream.
Read original ↗ -
22TechCrunch08:12
Tilly Norwood's press tour is going about as well as you'd expect for an AI
Interviews with the AI-generated performer Tilly Norwood have gone badly, including one in which the character appears to malfunction and starts speaking Chinese. The rollout is a visible test of whether synthetic personalities hold up in unscripted media settings.
Read original ↗ -
23TechCrunch23:30
Open or closed AI? Nvidia's Nader Khalil and Sydney Sykes take on one of the decisions shaping next-gen startups at TechCrunch Disrupt 2026
NVIDIA's Nader Khalil and Sydney Sykes will discuss whether startups should build on open or closed models on the Builders Stage at TechCrunch Disrupt 2026. The choice shapes cost structure, control over deployment and defensibility for early-stage AI companies.
Read original ↗
