Opinion | How Scared Should We Be of A.I. Right Now?
By Gammatek ISPL, Industrial Systems & Compliance Analyst at Gammatek ISPL
Last updated: September 2026 | 10 min read
Author block: Gammatek ISPL advises manufacturing, chemical, and pharma clients on risk and compliance frameworks at Gammatek ISPL, including how emerging technologies like AI get incorporated into existing risk management structures.
Why This Matters
Every few months, another wave of headlines asks the same question in a different tone: how scared should we be of AI? Depending on which piece you read, the honest answer ranges from "not at all" to "existentially." That range itself is the tell — when experts, executives, and columnists can't converge on even a rough consensus, the question is probably being asked the wrong way. If you run a business, manage a team, or sit on a risk committee, "how scared should I be" isn't actually a useful question to answer. The useful question is: what specific risks does AI introduce into my operation, and do I have a framework to manage them? That reframe is the entire difference between anxiety and preparedness — and it's the one the public conversation almost never makes.
Why "How Scared" Is the Wrong Question
Fear is not a risk category. It's an emotional response to uncertainty, and AI commentary right now has an unusually wide uncertainty band — genuine disagreement among credentialed experts about timelines, capabilities, and consequences. When public discourse tries to compress that uncertainty into a single "how worried should you be" dial, it collapses several genuinely distinct risk types into one undifferentiated anxiety:
Labor market disruption (a real, measurable, near-term risk)
Misuse by bad actors (fraud, disinformation, cyberattack automation — a real, present risk)
Systemic/existential risk from advanced future systems (a real but highly uncertain, long-horizon risk debated even among AI safety researchers)
Mundane operational risk (a system giving a wrong answer in a business process — the most common real-world AI failure, and the least discussed)
These four things have almost nothing in common except the label "AI risk." Any answer to "how scared should we be" that doesn't specify which of these it's addressing is close to meaningless.
What Risk Management Actually Does With Uncertainty Like This
This is where it's worth borrowing from a discipline that's dealt with exactly this kind of problem for decades: enterprise risk management. Risk managers don't ask "how scared should we be of fire, of cyberattack, of a key supplier going bankrupt?" They build frameworks that assign likelihood, severity, and mitigation cost to each specific risk, then allocate attention and budget accordingly — regardless of how emotionally loud any individual risk category feels in the news that week.
Applied to AI, this looks concretely different from the "existential dread vs. techno-optimism" framing dominating public discourse:
Public AI Discourse | Enterprise Risk Management Approach | |
Framing | "How worried should we be?" | "What's the likelihood and impact of this specific failure mode?" |
Unit of measurement | Emotional intensity | Probability × severity × cost of mitigation |
Time horizon | Often blended (near + far future) | Explicitly separated (quarterly, annual, multi-year risk registers) |
Decision output | An opinion | A specific control, policy, or monitoring requirement |
An organization using an actual enterprise risk management framework isn't asking "should we be scared of AI" — it's asking "what's our exposure if an AI-assisted process makes an undetected error in a regulated workflow, and what control do we have in place to catch it before it causes harm?" That's a question with an actual answer, and one you can act on.
A Concrete Example From Client Work
Placeholder structure to fill in:
What AI-assisted tool or process the client introduced
What specific failure mode was identified as a risk (not "AI is scary" — a concrete scenario)
What control or monitoring step was put in place to manage it
What changed in outcome or confidence level after that control existed
Implementation Considerations for Bringing AI Into an Existing Risk Framework
If you're responsible for risk, compliance, or operations at a plant or facility introducing AI-assisted tools, a few concrete steps that replace "how scared should we be" with something actionable:
Add AI-related failure modes to your existing risk register rather than treating AI as a separate, special category requiring its own entirely new framework. Most AI risks map onto risk types you already track (data integrity, third-party vendor risk, operational error, compliance documentation gaps).
Require a human verification step for any AI output feeding into a regulated decision — audit trails, safety documentation, and compliance records should never rely solely on unverified automated output, regardless of how good a tool has demonstrated itself to be historically.
Set a review cadence for AI-assisted tools, not just a one-time approval. A tool that performed well in testing can drift as underlying data or operating conditions change — schedule the same kind of periodic reassessment you'd apply to any critical safety system.
Separate near-term operational risk conversations from long-horizon societal risk conversations in your own internal discussions. Conflating "will this monitoring tool miss a defect" with "will AI eventually surpass human intelligence" produces exactly the same muddled, unproductive anxiety the public discourse suffers from — and it doesn't help your team make better day-to-day decisions.
The Honest Answer
So, how scared should we actually be of AI right now? Scared enough to take specific, nameable risks seriously — and not scared in the undifferentiated, headline-driven way that produces anxiety without action. The organizations that will navigate this well aren't the ones with the strongest opinion on the "scared vs. not scared" spectrum — they're the ones that already have a risk framework capable of absorbing a new category of risk without needing to reinvent how they think from scratch.
How This Connects to Your Own Risk and Compliance Framework
If your plant or facility is introducing AI-assisted monitoring, maintenance, or compliance tools, the same principle applies: the technology's real risk profile is manageable once it's folded into a structured framework — with clear audit trails, human verification steps, and a defined review cadence — rather than treated as an ungoverned new category.
Comments