Lens explains how people can study the same evidence and reach different conclusions. It also shows which statements have strong support and which do not.
What should people and organisations do to make AI responsible?
Many governments agree that powerful AI needs rules. They disagree about who should be responsible, what information should be public and when a person should be able to step in.
The shared ground
The European Union has legal duties for some companies that provide AI models used for many purposes. The United States NIST framework is voluntary. Singapore focuses on practical controls, clear human responsibility and human oversight when organisations use AI agents.
Supported by source Official sources are shown below.
Brussels
“
Large AI systems need clear records and clear responsibility.
One AI model can affect millions of people through products its maker never sees. Companies should record how a model was made, explain its limits and report serious problems. This helps people find who was responsible when something goes wrong.
What this view explains well
Risk can grow when one system is used by many people.
What this view may miss
Smaller companies may struggle with the cost of these duties.
Where that leaves us
Most people agree that someone must be responsible for harm caused by AI. They disagree about who that should be.
Should it be the company that made the AI model, the organisation that uses it, the government or the person operating it? Different groups answer differently because they face different risks.
Worth discussing?
Share the evidence and the different views.
Evidence map
Check the evidence yourself.
Lens separates information supported by an original source from our explanation. Open each source and decide what it supports.
Supported by source
European Union rules require some AI model providers to keep technical records, give information to companies that use their models, follow copyright law and publish a summary of training content.
Singapore’s guidance asks organisations to limit what AI agents can do, assign clear human responsibility, add technical controls, test and monitor systems, and explain responsibilities to users.
The same rule can protect the public and also be harder for a small company to follow. Which effect matters more depends on who carries the cost and the risk.
Lens
How this issue was made Three official sources reviewed · source material kept separate from our explanation · Lens does not promise neutrality or perfect truth · published 13 August 2026
Founding membership
One issue. Every week. Several clear views.
Founding members receive the full weekly Lens, the evidence map and a vote on which views we explain next.
One important issue with clear sources
Three to five meaningfully different views
Source material, claims, explanations and unknowns kept separate
Cancel any time
Founding price
A$12/ year
The price equals A$1 a month and is billed annually. It stays the same while you remain a member.