Trust, But Verify: What Chip Design Teaches Malaysian Businesses About AI Risk
Semiconductor Engineering argues that AI cannot be trusted the way human engineers are — and the chip industry's verification culture is the governance template every AI adopter needs.

Semiconductor Engineering, a specialist publication for chip design engineers, has published a piece titled "Trust, But Verify" with a blunt thesis: the chip industry trusts human designers to do their best while accepting they cannot be perfect, which is exactly why rigorous verification exists — and when AI enters the workflow, it "cannot be trusted" in the same way. The implication for Malaysian businesses deploying AI automation is practical, not philosophical. Before you let any AI system touch customer-facing work, financial decisions, or personal data, you need a defined way to check its output — the same sign-off discipline chipmakers have used for decades.
AI Summary
Semiconductor Engineering, a specialist publication for chip design engineers, has published a piece titled "Trust, But Verify" with a blunt thesis: the chip industry trusts human designers to do their best while accepting they cannot be perfect, which is exactly why rigorous verification exists — and when AI enters the workflow, it "cannot be trusted" in the same way. The implication for Malaysian businesses deploying AI automation is practical, not philosophical. Before you let any AI system touch customer-facing work, financial decisions, or personal data, you need a defined way to check its output — the same sign-off discipline chipmakers have used for decades.
Key Takeaways
- The semiconductor industry never takes even its best engineers on faith; every design is independently checked through verification before it ships. This is a culture, not an afterthought.
- The core claim of the piece: human designers earn calibrated trust because their failure modes are familiar. AI's failure modes are not — so trust must be replaced with testing.
- Verification skill is a transferable asset. As AI spreads into design, code, and operations work, the people who can independently check AI output become the bottleneck — and the opportunity.
- For Malaysia's E&E corridor around Penang and Kulim, verification engineering is an existing strength that maps directly onto the emerging job of AI assurance.
- Agentic AI raises the stakes: an agent that acts, not just answers, needs verification gates built between its decisions and the real world.
What Happened
Semiconductor Engineering published an opinion and analysis piece titled "Trust, But Verify." The argument, in the publication's own framing, is compact: "We trust designers to do the best they can, but know they cannot be perfect. That's why we verify. When AI gets involved, it cannot be trusted."
To understand why that short statement carries weight, you need to know what verification means in chip design. A modern chip contains billions of transistors and enormous complexity. No human team, however good, can hold all of it in their heads or get every detail right on the first
Sources & References
AIBlog summarises and analyses published information. We do not reproduce full source text. Analysis is editorial and not financial or legal advice.


