Breaking

Coinbase CEO Says AI Will Make Software Safer by Favoring Defenders

June 30, 2026 at 21:40 EDT

Security
Software Dev & Coding
Sales & Business

Coinbase CEO Brian Armstrong said on June 30, 2026 that AI will make software more secure, giving defenders an edge over attackers because they can scan all code before it reaches production. The remark pushes back against the more common view that AI introduces new vulnerabilities.

June 30, 2026 · Coinbase

"AI Will Make Software More Secure, Not Less"

Coinbase's CEO argues AI favors defenders over attackers — they can scan all code before production. But every AI-written line still gets 100% human review, and industry studies stay skeptical.

~40%

of daily code is AI-generated
(earlier target: over 50%)

~55%

of U.S. fraud cases handled with AI
(each outcome human-validated)

100%

of AI code reviewed by humans
"no vibe coding to production"

The claim vs. the studies

Share of AI-generated code found insecure in independent testing.

45%

Veracode
failed security tests

35–40%+

Other studies
SQL injection, XSS & more

100%

Coinbase safeguard
human-reviewed code

Coinbase's containment loop

AI writes code
Cursor · Copilot · Claude

→

Human reviews every line
no direct-to-production

→

Track metrics monthly
bugs · incidents · lead time

Armstrong's case

AI "favors defenders over attackers" — they can scan all code before production. Enforced reviews let productivity gains land without sacrificing quality.

Industry caution

A large share of AI code carries vulnerabilities. Risks rise when developers over-trust outputs and iterative prompting worsens flaws.

The point of agreement: for now, AI is not trusted to make autonomous production decisions in security-critical areas — human oversight stays essential.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.