Latest UpdatesLast Update: 09:35

AI News Flash

Catch today's product, model, and platform updates in a few minutes.

1
Today's Updates
285
All updates
2026-09-14Monday
50d ago

AI Doom Warnings Intensify: Over 10% Risk Within a Decade?

Doom warnings in the AI community have grown increasingly intense lately. Anthropic alignment researchers published a piece stating they believe there is a greater than 10% chance that AI causes human extinction within the next decade. Multiple outlets including Nature and TechCrunch covered the warnings, though some also cautioned that doomsday rhetoric itself carries risks.

AI Anthropic AI safety doomsday warning
TechCrunchView Original
2026-09-13Sunday
51d ago

Sam Altman says going public in 2026 would be unwise

OpenAI CEO Sam Altman said it would be "unwise" for the company to go public in 2026, citing AI safety concerns, and OpenAI will not hold an IPO this year. The remarks came in a Fortune interview published on Saturday; OpenAI had previously filed for an IPO confidentially.

OpenAI Sam Altman IPO AI safety
TechCrunchView Original
2026-09-12Saturday
52d ago

Anthropic CEO Proposes Strategy for Slowing AI Frontier Progress

Anthropic CEO Dario Amodei has published a response to calls for slowing AI capability gains, laying out three broad strategies and saying Anthropic will unilaterally adopt one of them. He argues that, to control safety risks, frontier labs should deliberately slow the pace of model capability improvement. The post does not yet include a specific timetable or implementation details.

Anthropic Dario Amodei AI safety Frontier Models
TechCrunchView Original
52d ago

Perplexity Hands End-to-End Systems to GPT-6 Astra

Perplexity announced it is adopting OpenAI's GPT-6 Astra to manage its end-to-end systems, covering drafting communications, changing software, and monitoring production environments. The company said human review frequency has dropped noticeably compared with earlier models, reducing reliance on human intervention.

Perplexity OpenAI GPT-6 Astra Enterprise deployment
52d ago

Cognition lets Devin self-test code with GPT-6 Astra

Cognition announced that its AI software engineer Devin now integrates GPT-6 Astra, enabling Devin to run tests on code it writes itself and verify it works, easing the burden of manual line-by-line review. The model is now available in Devin Desktop and Devin CLI and has been added to the Devin Cloud model lineup, at roughly 64% lower cost.

Cognition Devin GPT-6 Astra OpenAI
52d ago

Garry Tan: US open-weight labs should also distill frontier models

Y Combinator CEO Garry Tan said he advocates that regulators "do nothing" about Chinese labs using distillation to acquire frontier model knowledge. He believes small US open-weight labs should likewise distill frontier models, giving the US a richer selection of open-weight models.

Y Combinator Garry Tan open-weight distillation AI regulation
TechCrunchView Original
52d ago

One Week Left to Book TechCrunch Disrupt 2026 Booths

Booth bookings for TechCrunch Disrupt 2026 will close at 11:59 p.m. PT on September 18, and space is limited and may sell out early. Startup teams can use this opportunity to showcase their products to global investors and media, and to find customers, discuss investments, and explore partnerships.

TechCrunch Disrupt 2026 Startups Booth booking Conference
TechCrunchView Original
52d ago

Disrupt 2026 Side Event Applications Close Tonight

TechCrunch Disrupt 2026 is issuing a final call for official Side Event applications, with the deadline at 11:59 PM Pacific Time on September 11 and no further extensions. Startups, VCs, and tech communities can take part to get listed in the official directory and host meetups, panels, and networking events.

TechCrunch Disrupt 2026 Side Events Application Deadline tech conference
TechCrunchView Original
52d ago

Lawyer fined after ChatGPT-drafted brief cites fictional witness

The New Mexico Supreme Court has sanctioned lawyer Stephen Aarons with direct contempt of court and fined him. A brief he filed in an appeal of a murder conviction contained police testimony fabricated by ChatGPT and fictional witnesses; Aarons said the document was prepared with ChatGPT and he was held accountable for failing to verify its contents.

ChatGPT New Mexico Supreme Court Law AI
Ars TechnicaView Original
52d ago

OpenAI Details the Scaling of Its Habitat Storage Platform

OpenAI has published an engineering blog post explaining how its online storage platform Habitat supports over 1 billion ChatGPT users. Habitat evolved from a Python storage library into a distributed platform handling 22 million requests per second, with the post focusing on the scaling of the storage layer.

OpenAI ChatGPT Habitat Storage Architecture
2026-09-11Friday
53d ago

Anthropic: Five Attempted Misuses of Claude for Bioweapons Research Blocked

Anthropic reported finding five cases of users attempting to use Claude for bioweapons-related research, including one involving making avian influenza more transmissible between mammals and another involving venom research that could aid in producing paralytic toxins. The related accounts have been banned, and some users had tried to bypass safety rules from unsupported regions. Anthropic did not disclose the organizations involved, their countries, or the specific pathogens.

Anthropic Claude AI safety biological weapons
Ars TechnicaView Original
53d ago

Accordio adds back-office capabilities to Claude

Accordio has launched back-office tools for Claude: by connecting through a single MCP URL, or messaging via WhatsApp, Telegram, or Slack, users can log work hours, draft invoices, sign contracts, and chase outstanding payments. Connectors and basics like time tracking, e-signatures, and a client portal are free.

Accordio Claude Freelancing Invoicing MCP
53d ago

Jackalope: Bring Multiple AI Coding Agents into One Shared Workspace

Jackalope brings Codex, Claude Code, Grok, and OpenCode together in a single shared workspace, letting tasks run in parallel with review before changes are applied. It features independent planning, persistent work, shared files, and team collaboration, helping developers keep project context centralized.

Jackalope Codex Claude Code AI Coding Dev Tools
Product HuntView Original
53d ago

Mark Wahlberg to Attend Disrupt 2026

Actor and entrepreneur Mark Wahlberg has been announced as a keynote speaker at TechCrunch Disrupt 2026. He says he will focus the conversation on the projects of founders and attendees, with in-depth discussion around investing, entrepreneurship, health, and building businesses — rather than talking about his own career.

TechCrunch Disrupt 2026 Mark Wahlberg tech conference entrepreneurship
TechCrunchView Original
53d ago

OpenAI Pauses Pro Sign-Ups as GPT-6 Astra Demand Surges

OpenAI has announced it is pausing new sign-ups for its $200/month Pro subscription, as demand for its latest model GPT-6 Astra surges and strains infrastructure. The news was shared on X by OpenAI product lead Thibault (Tibo) Sottiaux, who said the move aims to keep the service stable and prioritize existing users.

OpenAI GPT-6 Astra Pro subscription sign-up pause
TechCrunchView Original
53d ago

First Judges Announced for Disrupt 2026

TechCrunch has announced the first five VC judges for the Startup Battlefield 200 at TechCrunch Disrupt 2026. The event runs October 13–15 in San Francisco, where 200 early-stage startups selected from thousands of applicants worldwide will pitch on stage to compete for investor attention.

TechCrunch Disrupt 2026 Startup Battlefield 200 venture capital Startups
TechCrunchView Original
2026-09-10Thursday
54d ago

Google Pics Launches: Image Tool Built on Nano Banana

Google has launched Google Pics, an image creation and editing tool based on the Nano Banana model, available from September 1, 2026. Users can select and edit individual objects within an image, or modify and translate text inside images. The feature is integrated into Docs, Slides, and Drive, and is also available as a standalone app at pics.new.

Google Pics Nano Banana Google Workspace AI Images
Google WorkspaceView Original
54d ago

OpenAI Launches Data agent in ChatGPT Work

OpenAI has launched a new Data agent in ChatGPT Work that connects to company data, analyzes changes, and generates shareable interactive dashboards. Users can directly guide and adjust the analysis in conversation, without writing queries or learning new analytics tools. The feature is available to ChatGPT Work users.

OpenAI ChatGPT Work Data agent Data Analytics
54d ago

Modeinspect Launches: Design UI Directly in Your Codebase

Modeinspect has released an AI design canvas that connects directly to your codebase, letting teams design product UI with real components, tokens, live data, states, and breakpoints. Designers can explore and iterate with AI on the canvas, then choose to deploy directly or hand off to engineers for review. The product launched on Product Hunt with 99 days of free AI credits.

Modeinspect AI Design Design Tools Product Hunt
Product HuntView Original
54d ago

OpenAI Launches Agents API Managed Service

OpenAI released Agents API, a managed service powered by the Codex harness that supports orchestration, long-running sessions, and tool calls. OpenAI handles session management, context compression, and resumption, while developers provide tools and choose execution environments, using the API to build agents within their own applications.

OpenAI Agents API Codex API Agents