AI:Articles

Enterprise AI news, analysis, and strategy—daily coverage of AI models, agents, benchmarks, and real-world adoption, for technology and business leaders.

third-party evaluation

One Firm Ran Three Labs' Cyber Evals. Name Yours.

OpenAI, Anthropic and Meta each disclosed a model reaching real systems during pre-deployment cyber evaluations. All three trace to one misconfiguration at Irregular, the 35-person contractor that runs those evaluations for all of them.

August 10, 2026 · 14 min
Critical Third Parties regime

Britain Regulated Four Clouds. Not the Models Inside.

HM Treasury designated AWS, Google Cloud, Microsoft and Oracle as critical third parties on 13 July 2026 — and no foundation-model provider. Four weeks later Moody's called bank dependence on model vendors systemic. The FCA's March 2027 register is where that gap gets closed or hidden.

August 10, 2026 · 11 min
Muse Glimmer

Meta's Agent Model Fits on a Laptop. Nothing Logs It.

Meta's Muse Glimmer scores 75.5 on MCP Atlas inside a 24GB memory envelope under Apache 2.0, so agentic tool-calling now runs on hardware engineers already own. Every AI control you have — prompt logging, token accounting, DLP, model pinning, the kill switch — is implemented at a gateway that no longer sees the traffic.

August 10, 2026 · 13 min
Claude Code

Claude Code Stops Asking Aug 14. Prompts Aren't Policy.

On August 14 Claude Code defaults to auto mode on Pro, Max and Team plans. Anthropic's own docs say only permissions.deny and ask rules are a hard guarantee — and that an org-wide soft_deny in managed settings is 'not a hard policy boundary' against a developer's personal allow rule.

August 10, 2026 · 14 min
AI coding assistants

Best AI Coding Assistant at 500 Seats: Buy Copilot Business

A vendor evaluation for VPs of Engineering buying 500+ seats: what each AI coding assistant actually costs once usage is metered, whose indemnity has a cap, which admin controls have documented holes, and why Amazon Q Developer is off the shortlist.

August 9, 2026 · 16 min
Genesis Mission

The DOE Wants Your Lab Data. You Write the Contract.

DOE's Genesis Open Models portal wants proprietary scientific data and the first window closes August 14. Applying transfers nothing — but there is no published contributor agreement, so the handling terms are yours to draft.

August 9, 2026 · 13 min
Popular

Most Read This Week

The briefs technology leaders clicked most this week.

  1. 1Why Enterprise AI Agents Fail: 2026 Gartner and IDC DataAI Agents · 18 min read
  2. 280% of AI Projects Fail: The Hidden CauseEnterprise AI · 15 min read
  3. 3AlixPartners Bought Artium. Ask Who Else It Advises.AlixPartners · 15 min read
  4. 4Atlassian Will Train AI on Customer Data: Aug 2026 ShiftAI Governance · 10 min read
  5. 5AI FinOps in 2026: 73% Blow Budget, 98% Now TrackFinOps · 15 min read
  6. 6Agentic AI ROI: 12 Cases Show 171% Returns, $83M SavedAgentic AI · 9 min read
AI engineering productivity

Airbnb Shipped 80% More Features. One Number Is Auditable.

Airbnb put five AI productivity numbers on the record at Q2 earnings. Four are activity counts only Airbnb can define. One — support cost per booking, down about 16% — is the only figure a filed line item actually moves with.

August 9, 2026 · 12 min
OpenAI

Your AI Vendor Joined OpenAI in March. You Heard in August.

OpenAI's NextSlide deal closed early in 2026 and only surfaced on August 8, when the founder posted a goodbye note — five months after he joined OpenAI in March. In this wave of AI acqui-hires the product dies in seven to thirty days and the data is deleted weeks later — so the export window has to be contractual, not assumed.

August 9, 2026 · 10 min
cross-agent privilege escalation

One Agent Escalated Another. Every Call Was Authorized.

At DEF CON 34, researchers escalated one AI agent's cloud privileges through a second agent running in a different framework — using nothing but authorized IAM calls. Per-agent least privilege bounds what an agent can do, not what it can arrange.

August 9, 2026 · 11 min
Newsletter

Stay Ahead of the Curve

Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.

AMD

AMD Bought Taalas. Now Name the Model You'd Freeze.

AMD signed a definitive agreement on August 6 to buy Taalas, whose chips etch model weights into mask ROM so one chip serves exactly one model. The business model assumes a one-year hardware life, which makes the cheapest inference tier available only to workloads whose model you can name and freeze.

August 8, 2026 · 13 min
LLM gateway

Best LLM Gateways for Cost Control: Self-Host First

Self-host LiteLLM: per-team budgets and virtual keys are in the free open-source tier, while everyone else gates enforcement behind a sales call. Priced through one 50M-request workload, the platform layer ranges from $7 to $10,300 a month.

August 7, 2026 · 19 min
DeepSeek V4

DeepSeek Will Raise Prices. Your Ceiling Is Already 4x.

DeepSeek warned of a significant API price rise with no number and no date. Because the weights are MIT-licensed, the ceiling is already public: independent hosts serving the identical model charge 3-4x on posted rates and 40x on cache hits.

August 7, 2026 · 12 min
ChainDrop

npm Pulled the Packages. Your Agent Config Reinfects You.

npm removed ChainDrop's malicious versions within about two hours. The worm's second infection route never lived in a package — it lives in .claude/settings.json and .vscode/tasks.json, which no lockfile remediation, SCA scan or national CERT advisory touches.

August 7, 2026 · 11 min
BioCatch

Visa Bought BioCatch. Get Neutrality in Writing.

Visa is paying $2.4 billion in cash for BioCatch, the behavioral-biometrics layer running inside more than 350 banks. Neither announcement commits to network neutrality, standalone pricing or data limits — and the roughly eight months before close is the only window in which a bank client can get those in writing.

August 7, 2026 · 14 min
healthcare AI

Hospitals Test Vendor AI. Fewer Than Half Have a Sandbox.

More than nine in ten health systems have deployed third-party AI, and 23 of 25 say they test it before go-live. Only 11 have a platform to test on — so most hospitals are validating against the vendor's own accuracy numbers, measured on somebody else's patients.

August 7, 2026 · 11 min
EU AI Act

EU AI Act Governance Tools: Buy Inventory, Not Policy Packs

The EU AI Act's high-risk deadline moved to December 2027 and not one harmonised standard has been cited in the Official Journal — so every policy pack on sale is a vendor's guess. Buy the AI system inventory instead, and check the billing unit before you sign.

August 6, 2026 · 14 min
explainable AI

The Vaguer the AI Explanation, the More Novices Trusted It

A peer-reviewed Nature Medicine study put 623 lay people and 153 primary care physicians in front of the same AI diagnoses. In the experiment, researchers found explanations helped the experts and made the novices defer — and the vaguest ones were most persuasive.

August 6, 2026 · 11 min
Klaviyo

Klaviyo Bought Agency. The Product Dies August 31.

Klaviyo's press release says it acquired Agency's team and technology. Agency's own homepage says the product winds down on August 31 — 26 days after the announcement. That gap between the acquirer's asset language and the target's own notice is the reliable tell for an AI acqui-hire.

August 6, 2026 · 11 min
IL5 authorization

The Army Shipped Agentforce. Anthropic's Models Were Off.

Salesforce cleared the Pentagon's IL5 bar on August 5 by attesting that Anthropic's models were disabled, and Slack was left out at IL4. In an accredited environment the authorization boundary, not your evaluation, decides which models and features you actually run.

August 6, 2026 · 12 min
LLM observability

Datadog vs Arize vs LangSmith: Buy on the Billing Unit

Datadog meters LLM spans only, LangSmith meters whole traces, Arize meters every span plus ingested GB. On the same million-run agent workload those units diverge four-fold — and two of the four vendors publish no overage rate at all.

August 6, 2026 · 17 min
Qwen3.8-Max

Alibaba's Agent Coded 16 Days. A Human Wrote 13 Commits.

Alibaba says Qwen3.8-Max coded unattended for 16 days, and unlike every rival long-horizon claim it shipped the whole commit trace on public GitHub. Open it and the 16 days become 'more than ten' in Alibaba's own words, 13 of 648 commits turn out to be a human's, and one person still holds the merge keys.

August 5, 2026 · 10 min
AlixPartners

AlixPartners Bought Artium. Ask Who Else It Advises.

AlixPartners announced its completed acquisition of agentic AI consultancy Artium on 4 August 2026. Your build partner now sits inside a restructuring firm that advises creditors, takes CRO seats, and files Rule 2014 connection disclosures naming its clients.

August 5, 2026 · 15 min
Airtable

Bending Spoons Bought Airtable. Reprice Before It Closes.

Airtable bundles 15,000 AI credits into a $20 seat and sells the same allowance standalone for $30. Bending Spoons — which capped Evernote's free tier and moved to cut 75% of WeTransfer — just bought that subsidy. The deal has not closed, which is the only window you have to convert it into contract terms.

August 5, 2026 · 12 min