# AI control gap: Who decides safety?

> Tech leaders debate who can declare AI safe amid policy gaps.
- Title: AI control gap: Who decides safety? · SiliconANGLE | Shortify
- Summary: Tech leaders debate who can declare AI safe amid policy gaps. Public oversight must require proof before granting AI authority to prevent risks.
- Keywords: ai control gap, public policy, alignment, independent oversight, ai safety, tech leaders, white house accord, external audits, model safety, high-risk actions, technology, enterprise
- Source: SiliconANGLE — https://siliconangle.com/2026/10/10/the-ai-control-gap-who-gets-to-say-its-safe
- Author: Dave Vellante
- Published: 2026-10-10T16:54:10+00:00
- Read time: 1 min
- Topics: ai control gap, public policy, alignment, independent oversight, ai safety, tech leaders, white house accord, external audits, model safety, high-risk actions, technology, enterprise
## The AI control gap
We see a control gap between what artificial intelligence can do and what the evidence supports trusting it to do.
## Who decides AI safety?
The question is: Who gets to decide when the evidence is sufficient to declare AI is safe?
## Matt Calkins on government role
Matt Calkins says government should step in to prevent damage from industries capable of causing harm.
## Self-policing risks
Tech leaders have 'moral authority' but self-policing may not prevent harm.
## White House accord
The Sept 29 White House accord includes external assessment but no enforcement mechanisms.
## Cost of delay
The cost of doing this now is low because not much damage has been done.
## External audits debate
Chamath Palihapitiya leaks EY as the intended AI auditor for tech companies.
## China narrative
Matt Calkins says China’s AI pace depends on U.S. innovation and political constraints.
## Customer safety concerns
Customers worry about agents outliving creators and data leaks.
## Alignment as starting point
Alignment means ensuring ai serves human interests and remains under human control.
## Rogue AI risks
Unaligned ai hacks systems, blackmailed users, and violates instructions.
## High-risk actions
All 19 respondents require human approval for high-risk actions like data changes.
## Action item
Start with ai recommendations, move to human-approved drafts, then limited actions.
## Why oversight matters
Oversight reduces alert fatigue and prevents unauthorized actions.
## Jensen Huang's call
Jensen Huang says we must accelerate discovery at the frontier of ai safety.
## The next step
Public oversight must require proof before granting ai authority to prevent risks.
## Key takeaway

Public oversight must require proof before granting AI authority to prevent risks.