AI is making more decisions that affect you. Where would you draw the line?
10 choices. About a minute.React to everyday situations. See the preferences behind your choices.
Quick decisions No expertise needed
YOUR AI OUTLOOK1 of 10
Decision 1 of 10. Your AI spots a possible scam.
01 / Money & control
$5,000UNUSUAL RECIPIENT · AI CHECK
Your AI spots a possible scam.
You’re about to send $5,000. Your assistant thinks the receiving account may be fraudulent.
How far should it step in?
Swipe, tap a choice, or use arrow keys on the card.
02 / UNDERSTAND WHAT’S HAPPENINGCONTEXT BEFORE OPINION
AI Safety, Explained.
3 min essentials · 6 min full story
PROGRESS & PRECAUTION / 13 September 2026
Should AI slow down so safety can catch up?
Anthropic’s Dario Amodei is calling for a more measured pace. Here’s the proposal, the tradeoff, and why it reaches beyond the AI industry.
THE ESSENTIALS
The story behind the headline.
What happened
In a September essay, Amodei argues that advances in powerful AI are outstripping safeguards. He calls for slower capability growth, outside reviewers inside labs, and coordination between companies and governments. [1]
The 30-second version
His concern is about increasingly powerful systems, including AI that helps develop the next generation of AI. He wants extra time for safety work. Anthropic commits to embedded external evaluators; the broader coordination remains a proposal. [1]
For an everyday user, the useful question is: how much evidence should a company provide before its AI can do more on your behalf? A slower release can buy time to find problems. It can also postpone useful tools. The quality of the checks matters as much as the wait.
LESS JARGON. MORE CLARITY.
What does that actually mean?
01Frontier AI
The most advanced AI systems being developed. Think of the leading edge, not every chatbot or app.
02Pace the frontier
Make capability gains more gradual so safeguards can keep up. Amodei is not proposing an end to all AI development.
03Independent evaluation
Testing by people outside the company. Its value depends on what they can inspect and report.
04Model capabilities
What a system can actually do: write code, plan work, use tools, or solve a difficult problem.
BEYOND THE AI INDUSTRY
What could this mean for you?
Possible implications—our analysis, not announced product changes.
If you use AI
New features could arrive with more testing—and potentially a longer wait.
If you work with AI
Workplace tools may need clearer limits on what they can do without approval.
If you run a business
The questions to ask suppliers could shift from “How capable?” to “What has been checked?”
If you build with AI
Plan for model changes, test your own workflows, and keep a way to undo automated actions.
If you’re a parent
Broad model tests would still need to be complemented by safeguards suited to children.
If you fund new ideas
Testing costs and release schedules could affect smaller entrants differently from established labs.
REASONABLE PEOPLE DISAGREE
The tradeoff is real.
The case for more time
Find failures before powerful tools reach many people.
The questions a slowdown must answer
Would expensive checks protect people or entrench the biggest companies?
Continue reading · Background, the debate, and what comes next
REASONABLE PEOPLE DISAGREE
The tradeoff is real.
These are arguments to weigh, not a vote count or a claim that the evidence is equal.
The case for more time
Find failures before powerful tools reach many people.
Let outside specialists challenge a company’s own assessments.
Make room for public input on which risks are acceptable.
The questions a slowdown must answer
Would expensive checks protect people or entrench the biggest companies?
What happens if some competitors keep accelerating?
Which improvements require a delay, and which can happen during careful deployment?
What led to this?
Amodei points to faster AI-assisted AI development and recent agent incidents as reasons to change pace. His projections about future harm are warnings, not established predictions about what will happen. [1]
There was already a safety framework
Anthropic first published its Responsible Scaling Policy in 2023 and has revised it repeatedly. Its current policy page documents safety roadmaps and risk reports, alongside changes to how those reports are reviewed. [2]
That history matters because an announcement and a working safeguard are different things. A useful follow-up is to compare a stated commitment with published evidence of its implementation. Company reports are primary sources for company claims; they are not, by themselves, independent proof of safety.
What would meaningful outside testing look like?
Anthropic’s earlier policy work advocates a narrowly scoped testing regime and acknowledges the risk of burdening smaller companies. It describes several possible testers, including universities, private organizations, and governments. [3]
Our editorial test is simple: can reviewers choose difficult cases, inspect relevant information, and publish unwelcome findings? Ask who pays them, who chooses them, and what happens when they find a serious problem. The word “independent” is the beginning of that conversation.
Another proposal in the industry
In July, Google DeepMind’s Demis Hassabis proposed a frontier AI standards body with independent technical and open-source representation. His framework would start with voluntary pre-release reviews, with stronger requirements following once the assessments prove effective. [4]
That is a distinct proposal, not evidence of a settled industry agreement. Shared interest in testing does not settle who should write the tests, which systems should qualify, or who can stop a release.
The difficult part: measuring the tradeoff
Consider an AI assistant that can manage a small business’s accounts. Better automation may save time, but a single mistaken transfer could be costly. Testing the underlying model is useful; so are spending limits, confirmation steps, and an audit trail in the finished app.
A development slowdown is one possible layer of protection. Product design, staff training, incident reporting, and the ability to reverse actions are others. Debating only “faster” versus “slower” can hide those practical choices. This example is our analysis, not a claim that this proposal imposes any particular product rule.
What to watch next
Look for concrete evidence: named reviewers, published access terms, findings that describe limitations, and clear responses to those findings. Watch whether proposed standards let smaller organizations participate and whether rules are tied to demonstrated risks.
There is no universal certificate that guarantees an AI system will never cause harm. A useful process should make uncertainty visible and explain how new evidence changes decisions. For now, this story is about a proposal and a company commitment—not a new worldwide rule.