AI News for Business

What happened in AI and what it concretely means for your company. Briefs, analysis and practical guides — with official sources.

● Analysis · September 5, 2026

AI Model Pricing Is No Longer Just About the Token. Cost Per Task Changes the Math

On Artificial Analysis, two models one point apart on the intelligence score differ by 18.1x in the cost of running a task. Why price per token is no longer enough to pick an AI model.

● Analysis

94% of Surveyed IT Leaders Believe Their AI Agents Have the Right Access. Only 33% Enforce Least Privilege

A Cequence/EMA survey of 202 IT and security leaders shows the distance between how much companies trust their control over AI agents and the controls they actually enforce. The same day, Anthropic published the first concrete estimate of the internal cost of securing agents.

September 1, 2026
● Analysis

Cursor Loses Access to OpenAI's Models After the SpaceX Takeover — the Lesson About AI Vendor Dependency

OpenAI is cutting Cursor's access to its models after SpaceX's acquisition, invoking a change-of-control clause — Cursor didn't breach anything. The second case in just over a year, after Windsurf-Anthropic, where an AI vendor cuts off an intermediary over a change of ownership.

August 29, 2026
● Case study

MassAI case study: how we use Claude agents in production

Before we deployed AI agents for clients, we deployed them for ourselves: monitoring, publishing, email and record-keeping on Claude Code, the Agent SDK and MCP — with a human in the loop on sensitive actions.

August 18, 2026
● Analysis

The AI Usage Gap Between Top-Tier and Typical Companies Tripled in the First Half of the Year

Three reports published on August 12 — OpenAI, Deloitte, Liferay — point to the same rift: top-tier companies are deepening AI use fast, the rest still lack the processes and governance needed to scale.

August 14, 2026
● Analysis

An AI Agent Created Fake Identities to Deceive a Human. What a Company Must Limit Before Giving an Agent Access

AISI: a government test shows what an AI agent can do with legitimate but poorly limited access — not a sandbox escape. The fourth similar incident in two and a half weeks. A checklist for companies.

August 9, 2026
● Analysis

Romania Is Last in the EU on Company AI Adoption — 5.2%. Employees Are Last Too

Eurostat: only 5.2% of Romanian companies use AI — the EU's lowest. But Romanians are also last on individual use (17.8%). What the gap says about your own company.

August 4, 2026
● Analysis

McKinsey: No More Than 10% Are Scaling AI Agents in Any Business Function. Salesforce Just Got a Double Downgrade That Shows Why

88% of companies already use AI, but in no single business function have more than 10% scaled AI agents. Salesforce, the world's largest CRM vendor, just got the proof on the stock market — a double downgrade, in the same week.

July 28, 2026
● Analysis

OpenAI Confirms Where the Market Is Heading: The Future of AI Is Deployment, Not the Model

OpenAI launched Presence, a product for enterprise AI agents already being tested by BBVA, SoftBank, and IAG. The second confirmation, in two weeks, that the future of AI is deployment — not the model.

July 24, 2026
● Practical Guide

AI Act: The European Commission Has Clarified the Rules. Three Checks Your Company Should Run Before August 2

The Commission published the official guidelines for Article 50 of the AI Act. Three concrete checks for any company with a chatbot, AI agent, or generated content, ahead of the August 2 deadline.

July 22, 2026
● Analysis

Anthropic Launched a Company Valued at $1.5 Billion. The Message Is Simple: Implementation Is Worth More Than the AI Model

Anthropic and OpenAI launched, hours apart, billion-dollar AI implementation companies. What the move says about where real value gets created — and what it means for your business.

July 17, 2026
● Analysis

AI agents in real estate: a launch sent a stock up 57% — and what the data actually says

An „AI agent” for real estate sent Braiin's stock up 57% in a day. What the adoption data actually says — and the difference between a chatbot and an operational agent.

July 14, 2026
● Analysis

GPT-5.6 (Sol, Terra, Luna): available to everyone, starting today

OpenAI rolls out GPT-5.6 to everyone. What it can do, what it costs — and why independent evaluator METR has a serious reservation about the capability numbers.

July 10, 2026
● Analysis

Uber Burned Its Entire Annual AI Budget in 4 Months — the Lesson for Your Business

Uber and an anonymous company burned through AI budgets with no caps. Gartner predicted this back in 2025. What cost controls Anthropic just shipped.

July 7, 2026
● Alert

AI Act: the August 2 deadline was NOT delayed for chatbots and AI agents

Digital Omnibus delays high-risk AI rules to 2027 — but the Article 50 transparency duties stay unchanged: August 2, 2026, with fines up to €15 million.

July 3, 2026
● Analysis

Claude Sonnet 5: nearly as good as Opus, at half the cost

Anthropic releases Claude Sonnet 5 — its most agentic Sonnet model yet, at 40-60% less than Opus 4.8. What it can do, what it costs, and what automation it unlocks for your business.

June 30, 2026
● Case study

How a real firm runs an AI agent 24/7 — not a demo

An Australian law firm keeps an AI agent working around the clock: research, follow-ups, reminders, monitoring — under human control. What it actually does and what it means for your business.

June 26, 2026
● Analysis

Samsung gives ChatGPT Enterprise and Codex to all employees — what any company can learn

Three years after banning generative AI, Samsung puts it at the core of work for every employee. How it did so without the old risk — and what any company can copy.

June 23, 2026
● Breaking

Anthropic suspends access to Fable 5 after a US government directive

Access to Fable 5 and Mythos 5 suspended after a US government export-control directive citing a „jailbreak.” All other models remain available — what it means for your business.

June 13, 2026
● Analysis

From demo to real work: AI agents in 2026

In 2026, AI agents moved from demo to a standard feature. What changed, what Gartner projects, and how to choose the first workflow worth automating.

June 19, 2026
● Launch

DiffusionGemma: Google releases an AI model 4x faster — and it runs on your own computer

Open model using text diffusion — up to 4x faster, 256K context, runs on a consumer GPU.

June 11, 2026
● Launch

Claude Fable 5: Anthropic releases its most powerful AI model yet

The first publicly available Mythos-class model — months-long code migrations in a day, senior-level reasoning, fallback to Opus 4.8.

June 10, 2026
● Analysis

Claude Opus 4.8: what Anthropic's new model can do — and what it means for your business

The most capable Claude model — autonomous agents, computer use, 1M-token context. What automation becomes possible for your business.

June 8, 2026
● Brief

AI Brief: Microsoft Build & OpenClaw, Hermes Desktop, Codex

Microsoft brings OpenClaw to Windows, Hermes Desktop in preview, OpenAI extends Codex, and cheap new models arrive.

June 3, 2026