Future AI
AI is a tool.
Organizations choose the objective, data, limits, and deployment. They remain responsible.
Position paper
Our position on AI.
Use it where measured savings exceed its load. Regulate by risk. Keep people accountable.
Short version
Read this first.
Updated July 26, 2026.
Benefits need proof.
Start with a baseline. Include the full cost. Measure the result.
Risk sets the rules.
Higher stakes require stronger testing, records, review, and human authority.
Where we stand
Deployment must be accountable.
Start where the current waste can be measured.
Disclose electricity, cooling, infrastructure, and local impact.
The deploying organization owns the outcome.
Contracts should require a baseline, method, records, and results.
Governance
Oversight should match risk.
A note-taking tool and a system that affects health, employment, or access should not face the same rules.
Disclose and review.
Set access controls, review outputs, and protect data.
Test and record.
Keep logs, monitor errors, and provide an appeal path.
Validate independently.
Keep human authority, clear liability, and an incident plan.
Energy
Energy is part of the cost.
Electricity, cooling, and local infrastructure belong in every net-benefit claim.
U.S. data center electricity use in 2023.
DOE projection for 2028.
Questions
What this position means.
Direct answers to the obvious objections.
Does this excuse unlimited data center growth?
No. Every claim must include electricity, cooling, infrastructure, and local effects. If the total cost exceeds the measured benefit, the project fails the test.
Who is responsible when harm occurs?
The organization that deploys the system. A vendor or model cannot absorb legal or operational responsibility.
What if the vendor controls the data?
Then the contract must require access to records, a before-and-after baseline, and an outside audit path.
What about rebound effects?
Measure total use and total output, not efficiency per task. A lower unit cost is not a net benefit if overall cost rises.
What about jobs and access?
This paper does not predict job totals. Any system that affects employment or access needs records, review, an appeal path, and human authority.
Is this campaign conflicted because it uses AI?
Yes. Clear Frameworks used AI tools to research and build the campaign. That use is disclosed, the sources are public, and the campaign applies the same measurement standard to itself.
The standard
Ask for the proof.
Before anyone calls a project efficient, ask for the baseline, total cost, and measured result.