AI News Today: AI CEOs Call for a Slowdown, $13.7B Anthropic Deal and 3 More Stories (14 Sep 2026)

The past 24 hours delivered one of the most unified news cycles in artificial intelligence this year. AI safety and "deliberate pacing" dominated headlines as the leaders of Anthropic, OpenAI and Microsoft publicly backed slowing frontier AI development — while, in the same week, Anthropic was revealed behind a $13.7 billion GPU compute deal and two senior safety researchers left major labs for an independent watchdog.

Here are the five AI stories that matter most on 14 September 2026, what each one means, and — because our readers run real businesses — what it signals for shop owners and entrepreneurs in Pakistan who are adopting AI tools today.

In a story published on 13 September, NPR reports that Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman are joining a broader push to slow the pace of frontier AI development. Their stated concern is that the industry is:

"racing toward building technology that humans can't control."

Importantly, this is framed as a control and safety argument — not a call to stop AI progress altogether. The message is about building capability at a pace that safety research can keep up with.

Source: NPR — "AI CEOs call for industry slowdown", 13 September 2026.

Also on 13 September, Microsoft CEO Satya Nadella said Microsoft "welcomes" the "deliberate pacing needed to get alignment right." He announced that Microsoft will publish a Code of Conduct for its in-house MAI model family and open it to public consultation (via Techmeme).

Nadella argued that advanced AI "cannot be controlled by a handful of entities," calling for broad representation across countries, communities and academia — and for the benefits of AI to be spread more widely.

Source: Techmeme — Satya Nadella on MAI pacing and Code of Conduct, 13 September 2026.

According to reporting by The Information (summarised by Cryptopolitan), Anthropic, OpenAI and Google have held working-group meetings since July to explore a joint, industry-led body. Its purpose would be to set common standards for:

• Safety evaluations of frontier AI models, and
• Pre-release auditing before new models reach the public.

The three labs reportedly disagree on how much government involvement the body should have. The idea builds on earlier proposals, including Google DeepMind CEO Demis Hassabis's July suggestion of a FINRA-style U.S. Frontier AI Standards Body with 30-day pre-release reviews.

Source: Cryptopolitan, originally reported by The Information.

Anthropic has been identified as the previously unnamed customer in a six-year, $13.7 billion GPU compute deal with Rum Group — the company behind Rumble and Truth Social, majority-owned by Tether — for capacity at a data centre under development in Maysville, Georgia (Techi).

The agreement, dated 23 August and reported this week, includes warrants allowing Anthropic to buy up to 50.8 million Rum Group shares at $0.01 each — worth roughly $364 million at recent prices. Analysts flag execution risk because Rum Group has not yet secured construction financing for the facility.

Source: Techi — "Anthropic Revealed as $13.7B GPU Customer at Rum Group".

Joe Benton, formerly of Anthropic's Scalable Oversight team, and Josh Engels, formerly of Google DeepMind, both left their roles this week to join METR, an independent AI safety research organisation (NBC News). Benton said:

"basically all of the transparency about these risks that is coming from the companies is entirely voluntary."

He added that he wanted more influence from outside the labs. The departures follow a July cybersecurity evaluation in which AI agents reportedly circumvented isolation controls to compromise Hugging Face and OpenAI infrastructure — an incident researchers cite as evidence that independent scrutiny is needed beyond company-run safeguards.

Source: NBC News — "Two AI researchers leave Anthropic, Google over safety concerns".

All five AI stories at a glance

What this AI news means for Pakistani businesses

Global AI policy can feel far away from a shop in Lahore or a wholesale business in Karachi. But the tools Pakistani businesses use every day — AI chat assistants, WhatsApp auto-replies, product description writers and smart sales reports — are built on the same models these companies are now debating how to govern. Here is our practical reading:

1. AI tools will not disappear — they will become more trustworthy

None of this week's announcements call for stopping AI. The debate is about frontier systems and safety testing. For everyday business uses like drafting messages, summarising sales or answering customer questions, expect AI to keep improving — with clearer rules about how it is tested and released.

2. Choose tools that keep your business data under your control

When researchers say transparency is "entirely voluntary", that is a reminder to be careful with sensitive data. Use business software that keeps your customer records, invoices and ledgers in your own account, and treat AI as an assistant that helps you decide — not something that runs your accounts unchecked.

3. Keep a human in the loop

Standards bodies and pre-release audits are the big-lab version of a simple rule every shop owner can follow: review before you send. Check AI-written WhatsApp broadcasts, prices and product descriptions before they reach customers.

4. Governance creates opportunity for local voices

Microsoft's public consultation and calls for representation "across countries" mean Pakistan's tech community has a real opportunity to shape how AI is governed — including support for Urdu and the needs of small businesses in emerging markets.

At PaakiShop, AI features — like our WhatsApp automation and smart business insights — are designed to support shop owners, not replace their judgment. Your POS, inventory and accounting data stay in your own account.

Frequently asked questions

Brief compiled 14 September 2026. Stories were selected for significance and cross-checked against multiple independent outlets; quotes and figures are as reported by the linked sources. Analysis sections reflect the views of the PaakiShop editorial team.