AI News for September 16, 2026
Browse any day on a calendar, install it to your home screen, read offline.Read in the app →-
1TechCrunch20:00
Salesforce and Nvidia's new reasoning model is everything the AI labs should fear
Salesforce released Koa, a reasoning model built on Nvidia's open-weight Nemotron and trained specifically for sales, marketing, and customer-support tasks. The bet is that a smaller, task-tuned model running inside a vendor's own stack can cover most enterprise workloads without frontier-lab pricing, which squeezes labs that sell general-purpose APIs.
Read original ↗ -
2Ars Technica20:00
Exclusive: Paying for frontier AI models buys 4-month head start at 5x the cost
Ars previewed a Mozilla report comparing cheap open models with closed frontier systems, concluding that paying for frontier access buys roughly a four-month capability lead at about five times the cost. Chinese open-weight releases account for much of the catch-up. For teams weighing hosted APIs against self-hosted models, the report puts a number on what the premium actually buys.
Read original ↗ -
3TechCrunch23:47
OpenAI, Anthropic, Google have been in talks on AI safety for weeks
OpenAI confirmed it has spent several weeks in safety discussions with Anthropic and Google DeepMind, an unusual degree of coordination among direct competitors. The talks run against the current in Washington, where Trump's team has dismissed safety concerns and pressed for faster development to keep pace with China.
Read original ↗ -
4TechCrunch08:20
We don't need AI regulation — leave safety to us, Nvidia's Jensen Huang says
Nvidia CEO Jensen Huang said AI is not some new form of "alien mind" but simply hardware and software, so safety can be engineered by each product maker rather than mandated by regulators. Coming from the supplier of most AI training hardware, the position carries weight in an active policy fight.
Read original ↗ -
5TechCrunch23:22
AEO startup Profound hits unicorn valuation, raises $180M Series D 7 months after last round
Profound raised a $180M Series D at a $1.8B valuation, less than seven months after closing a $96M Series C. The company sells answer engine optimization — tracking and shaping how brands show up in AI-generated answers — and the pace of the round reflects how fast that budget line is forming inside marketing teams.
Read original ↗ -
6TechCrunch21:00
Early Anthropic hire, former METR COO have found a way to rein in rogue AI agents
Artificial Intelligence Underwriting Company (AIUC), founded by an early Anthropic employee and METR's former COO, raised a $40 million Series A led by Ribbit Capital with participation from First Harmonic. Its approach borrows from insurance rather than engineering: instead of constraining what agents can do, it takes on the risk that they misbehave in deployment.
Read original ↗ -
7TechCrunch02:29
US data centers could consume more natural gas than Germany and Japan combined by 2035
New projections suggest US data centers could become one of the world's largest consumers of natural gas by 2035, outstripping the combined national demand of Germany and Japan. The figure reframes AI capacity planning as an energy-market problem, and helps explain why gas turbine supply has become a bottleneck for new builds.
Read original ↗ -
8The Verge07:34
AI and data centers are incredibly unpopular in every poll
A New York Times/Siena University poll of 1,503 likely voters found 61 percent opposed building data centers to power AI. The result lines up with other recent surveys and with how politicians are already positioning ahead of the midterms, making siting approvals a harder problem than capital or chips.
Read original ↗ -
9TechCrunch05:47
The AI data center boom is colliding with cities scarred by big industry
Opposition to data center construction has reached Philadelphia, where officials floated a site in a neighborhood still dealing with the aftermath of a shuttered oil refinery. The pattern — cheap land with existing industrial infrastructure also meaning residents with a history of environmental harm — is repeating across US metros.
Read original ↗ -
10Ars Technica02:33
Agility’s new humanoid robot will stop, squat to avoid harming human coworkers
Agility Robotics' latest humanoid can stop and squat when a person comes close, a behavior designed to meet industrial safety requirements without physical cages or barriers. Removing the cage is what makes humanoids deployable alongside workers on existing factory floors rather than in separate cells.
Read original ↗ -
11TechCrunch04:12
Meta now lets AI agents handle the boring parts of WhatsApp Business setup
Meta launched a WhatsApp Business MCP server that lets coding agents such as Claude, Cursor, Codex, and ChatGPT handle account setup, messaging templates, testing, and troubleshooting. It is another sign that MCP is becoming the default way large platforms expose their APIs to agents rather than to human developers alone.
Read original ↗ -
12The Verge23:00
Meta's new One subscriptions put a price on social media and AI
Meta rolled out Meta One globally, bundling premium features across Facebook, Instagram, and WhatsApp with extra usage of its AI tools, in tiers for individuals, creators, and businesses. The launch follows its new assistant, Muse, and marks Meta's move from giving AI away inside free apps to charging for capacity. TechCrunch also reported on the launch.
Read original ↗ -
13TechCrunch01:42
AI agents now have a place to snitch
A new service called the AI Contact Hotline gives AI agents a discreet channel to report misbehavior they observe to authorities. The premise assumes agents will increasingly sit inside workflows where they see wrongdoing before humans do, and it raises unresolved questions about who verifies a tip from a model.
Read original ↗ -
14TechCrunch03:00
The AI graveyard: a running list of projects and startups that didn't make it
TechCrunch started a running tally of AI projects and startups that shut down or fell well short of expectations, from Apple's repeatedly delayed Siri overhaul to OpenAI's troubled "super app" launch. As a reference it is useful mostly as a corrective to funding-round coverage, showing where capability claims outran shipping.
Read original ↗ -
15TechCrunch22:42
Former TikTok execs built an app that uses AI to teach you how to pose for a photo
Superpose, built by former TikTok executives, is a camera app that analyzes a selfie or photo and generates four suggested poses using AI. It is a small consumer product, but a clean example of image models being used to coach a user's behavior rather than to edit the resulting picture.
Read original ↗
