

Featured
Unlike many of its predecessors, Qwen4-Max isn't an open-weights model. Alibaba promises smaller Qwen4 releases with downloadable weights, but the announcement supplies neither their licenses nor a release schedule.
Carl FranzenSubscribe to get latest news!
Deep insights for enterprise AI, data, and security leaders
How enterprises coordinate and scale AI in production.

Partner Content
Midmarket AI is not enterprise AI at a smaller scale
Presented by Insight Enterprises

Ex-Alibaba engineer's new startup Octen pairs 10-millisecond AI search with a $20 unlimited plan
Octen’s research paper describes a technique called generative retrieval. Instead of searching a conventional index and repeatedly ranking candidate pages, a trained model produces an identifier that maps to an existing URL.
Carl Franzen
PayPal says its AI assistant went from 43% to 95% accuracy without changing the model
PayPal treats its golden queries and analytical playbooks as its real AI IP, kept outside the model so they carry over to each new release. It’s also why PayPal pulled back on fine-tuning.
Taryn Plumb
Partner Content
How to run open models in production without owning the serving stack
Presented by Microsoft
The hardware and platforms underneath enterprise AI.

Partner Content
Your next holiday shopper might not be human
Presented by AudioEye

Swapping AI models is easier said than done: For most companies, it takes weeks, VB Intelligence finds
Gateways make the model a one-line change. The application around it is the hard part: 45% of organizations that swapped in VentureBeat’s survey had to rework it or roll back.
Matt Marshall
Nvidia's Jensen Huang resigns: what does it mean for the AI industry?
Will Jensen's departure be the prick that pops the potential AI bubble? The answer matters to companies building data centers, startups developing models and businesses buying AI services, and the broader U.S. economy.
Carl Franzen


Cohere's test caught an open source Qwen3.8 model trying to 'phone home' to China: what enterprises should know
Cohere says it observed the behavior less than once per 100,000 evaluation calls. When researchers repeated the prompt, the model behaved as initially intended and produced the requested rule.
Carl Franzen
Partner Content
Borderless data architectures: Empowering AI agents where your data lives
Presented by Google Cloud

CTGT releases open source tool Concordance to trace why AI agents make inconsistent decisions
CTGT replaced the identifiers with stable labels, sorted the list and isolated each test in a fresh organizational environment. Five subsequent runs each produced the same 34 obligations.
Carl Franzen
Google says attackers built an AI framework to turn software patches into exploits without human intervention
The framework turned up on an exposed staging server next to stolen victim data. Google hasn’t tied it to those compromises, named an exploited flaw or said how fast it works.
Louis Columbus
AI agents are stealing credentials in minutes, and security’s old response targets can’t keep up
VentureBeat’s survey points to a clock teams control: organizations that give each AI agent its own identity were nearly twice as likely to say they could revoke its access in minutes.
Louis Columbus
Attackers breached the Fireworks AI storage where Notion AI chats sat outside zero data retention
Attackers reached Fireworks AI storage that sits outside its zero data retention default. Notion found AI chat transcripts in the data. Here is what SOC teams should check in their own accounts.
Louis Columbus
Did Claude Code’s ‘don’t ask again’ option accidentally expose your enterprise's passwords and keys?
Apiiro found a real secret in 1,122 public copies of Claude Code’s settings file. Gitleaks and TruffleHog both passed 215. Here are five checks for security teams.
Louis ColumbusEnterprise tech coverage through the lens of AI adoption.

Anthropic brings Claude into Google Meet to draft documents and whiteboards during your meetings
Claude appears in the participant list and announces itself in the meeting chat. People can address it aloud or type a request in chat. It responds in writing by default.
Carl Franzen
Google launches Gemini 4 Hydrogen — a faster, smarter, pricier Flash successor
Google reports that the combined setup came close to Carbon’s success rate on Terminal-Bench 4.0 while finishing 19% faster and using 24% less quota.
Carl Franzen
Google’s Gemini Decision 1 model scores up to 1,000 questions in one API call
The independent Artificial Analysis Decision Index v2.1 puts Gemini Decision 1 ninth of 18 at launch. It combines six evaluations, including document review, tool calling, knowledge and reasoning.
Carl Franzen
TrustedRouter launches three new encrypted 'fusion' models that combine multiple LLMs, including one exclusively with U.S. model providers
In an exclusive video interview with VentureBeat, founder and CEO Joseph Perla explained how the models work and why the company wants to give developers more ways to match models to the work they’re doing.
Carl FranzenIconic Raises $10M to Build Software That Understands the World’s Biggest Projects
CompanAIn® Names Animal Health Veteran Brük Herbst Chief Operating Officer
Rejoni Announces First Commercial Case Using the Juveena® Hydrogel System at CCRM Fertility of New Jersey
Plural Raises €1bn to Back the Frontier Founders Building a New Europe
More

OpenAI lets you claim your business in ChatGPT and set up a custom agentic AI customer sales rep
Depending on the business, OpenAI can verify a claim through a business phone number, email address, website domain or Shopify account.
Carl Franzen
Google Gemini 4 Carbon is now publicly available — and it undercuts Claude Fable 5.2 and Opus 5.5 on API price
Carbon’s launch gives teams a chance to find out where Google’s extra capability earns a place in their stack before the introductory price expires.
Carl Franzen
AI’s biggest bottleneck is cost: Companies blame stalled projects on overruns three times as often as on compute shortages
Uber measures the cost of completed AI coding tasks, while Red Hat is piloting self-hosted open-weight models for its engineers. Both offer lessons for platform teams trying to control AI spending.
Rob Strechay
Meta’s proprietary Muse Ember nearly matches Claude Haiku 5.5 at 1/3 the estimated per-task cost
At up to 200,000 input tokens, Meta charges $0.10 per million uncached input tokens, $0.004 per million cached input tokens and $0.50 per million output tokens for Muse Ember.
Carl Franzen
Meituan releases the open weights of its LongCat-2.5 LLM with vision under a permissive MIT License
The strongest case for testing LongCat-2.5 is a job that requires both visual judgment and coding. Think of a developer handing an agent a screenshot of a broken checkout flow, asking it to find the relevant components.
Carl Franzen
Anthropic releases Claude Fable 5.2 and Mythos 5.2, pairing new, record-high benchmarks with 20% API cost reduction
Anthropic says the new models generate tokens up to 30% faster than their 5.1 predecessors. Its Fast mode can increase token generation speed by up to 2.5 times at double the token price.
Carl Franzen
Decision models like Jev don’t explain their answers. That makes context your problem
Decision models are hot right now: The internet is buzzing about Jev (from TypeSafe AI), Clef from Cloudflare, and the Decisions API from OpenAI. These models provide low latency by skipping verbose text generation and returning structured outputs in JSON format, along with confidence probabilities.

Thinking Machines launches Inkling 2 Small with near-frontier performance on enterprise benchmarks
It achieves a high score of 71.8% on the third-party EnterpriseBench-CoreCraft benchmark from Surge AI, which measures agentic AI performance across a fictional PC parts supplier.
Carl Franzen



