The SI safety accord and your ERP: oversight, audit and a human on every irreversible action

What the voluntary SI safety accord asks of labs, why self-policing is not enough for software that touches money, and five checks for any SI vendor.

By Magathi··3 min read

Alongside the rename, six major technology leaders signed a voluntary Super Intelligence accord built on self-policing and safety checks. Critics call it symbolic. Either way, it is a useful lens for judging the SI you let near your money.

What happened

On 29 September 2026 the US government signed Executive Order 14365, "Inaugurating the Era of Super Intelligence", directing federal agencies to use "Super Intelligence" (SI) in place of "artificial intelligence" (AI). The same day, six technology leaders (Google's Sundar Pichai, Nvidia's Jensen Huang, Meta's Mark Zuckerberg, OpenAI's Greg Brockman, Anthropic's Dario Amodei and Elon Musk) signed a separate, voluntary safety accord some called a "constitution" for the technology.

This is a change of terminology inside the US federal government, not a law forcing every company to drop the word "AI", and "SI" covers the same technology the statute already called AI. The accord is voluntary and non-binding, and critics note it creates no mandatory oversight.

Why self-policing is not enough for business software

An accord between labs governs how models are built and tested. It says nothing about what a particular business system lets a model do with your bank account. That is decided by the software: its permissions, its approval rules and its audit trail.

Five checks for any SI vendor

  1. 1Does it keep a person on every irreversible action, or can the model act alone?
  2. 2Is there an append-only audit trail that covers the agent, not just people?
  3. 3Can the agent do more than the person who switched it on, or exactly the same?
  4. 4Does every number link back to a record, or does the model do arithmetic you cannot trace?
  5. 5Can you run it under the model of your choice, and switch if one is withdrawn?

A person on every irreversible action

In Omni, agents run in one of three modes (Only when I ask, Prepare it, ask me, or Autopilot) and there is no fourth. Financial actions are always shown to a person field by field; irreversible ones also need typing CONFIRM and are never automated. Every action, by a person or an agent, is written to an append-only audit log in the same transaction as the change.

See it on your own numbers

To judge SI with oversight, bring one real workflow (a month-end, a collections list, a stock count) and watch it run in Omni. Book a 30-minute call or message us on WhatsApp.

Frequently asked questions

Is the accord legally binding?
No. It is voluntary and creates no mandatory oversight.
Can an Omni agent act irreversibly in Autopilot?
No. Irreversible actions are never automated, in any mode.
Super IntelligenceArticleERP
M
Magathi

AI Visibility specialist at Edthi · SEO, AEO & GEO

Leads AI Visibility programs at Edthi, helping companies get found and recommended across Google and the AI engines. Writes about how AI search actually decides who to cite.

Ready when you are.

Book a 30-minute call, message us on WhatsApp, or drop your details and we'll reach out. Either way you'll see Omni run on a real workflow — you approve anything that moves money.

Talk to us about Omni

Tell us a little and we'll show you the agent running on a workflow of yours.

No spam. We reply within one business day.