Human Oversight and Company Policies

You are the Deployer

Bridging the Gap

In this final lesson, we move from understanding AI to professional accountability. Under the EU AI Act, you are classified as a deployer.

This means you are legally responsible for the outputs generated by the AI tools you use. Human oversight is not just a suggestion; it is your legal safeguard.

Welcome to the final stage of your AI literacy training. Today, we bridge the gap between AI capabilities and your professional accountability. Under the EU AI Act, you are officially a 'deployer'. This means you are the person responsible for ensuring AI remains a tool for productivity, rather than a liability for our company.

• You are a 'deployer' under the EU AI Act
• Human oversight is a legal and ethical mandate
• Accountability rests with the human, not the tool

The Human-in-the-Loop Mandate

Article 14 Compliance

The EU AI Act (Article 14) requires meaningful human oversight to prevent automation bias—the tendency to trust automated systems even when they are clearly wrong.

• Verification: Fact-check every claim.
• Intervention: You must be able to override the AI.
• Responsibility: You are the final filter.

The 'Human-in-the-Loop' mandate is a core legal requirement. It is designed to fight automation bias—that dangerous habit of clicking 'copy-paste' without thinking. You must act as the final filter. This involves three steps: verification of facts, the authority to intervene and change the output, and accepting full responsibility for the final message sent.

• Meaningful human oversight (Article 14)
• Avoid automation bias
• Authority to override AI suggestions

Aligning with Company Policies

The Three Pillars of Compliance

Our internal policy is built on three non-negotiable pillars:

• 🛡️ Data Privacy: Never input sensitive client data or PII into public tools.
• ⚖️ Bias Monitoring: Check for stereotypes in recruitment or performance summaries.
• 📜 Copyright: Ensure AI code or images don't violate IP rights.

To stay compliant, your daily usage must align with our core pillars. First, Data Privacy: never paste trade secrets or client names into public tools like ChatGPT. Second, Bias Monitoring: look for language that might unfairly exclude groups. And finally, Copyright: ensure generated content doesn't infringe on third-party rights.

• Data Privacy (No PII in public AI)
• Bias Monitoring (Inclusion and neutrality)
• Copyright & Licensing (IP protection)

Scenario: The Project Proposal

Interactive Review

You've used AI to draft a project proposal. Before sending it to the client, you must apply oversight.

Identify the two critical errors in the draft to proceed.

Imagine you've drafted a proposal using an AI assistant. It looks professional, but is it compliant? Review the document and click on the two areas that represent a compliance risk. Excellent work. Only after these manual edits is the document ready to be sent. This is meaningful human oversight in action. Well spotted. The AI accidentally pulled a specific client name from an old internal document. That's a major privacy breach if sent. Correct. The AI hallucinated a 10% discount that doesn't exist. You must manually correct the budget.

• Identifying hallucinations
• Preventing data leakage
• Applying manual corrections

The 4-Step Oversight Checklist

Your Daily Protocol

Use this checklist before every AI-assisted publication:

1. Accuracy: Verified facts and calculations?
2. Privacy: Removed all internal data?
3. Bias: Is the output inclusive?
4. Tone: Does it sound professional?

To make oversight easier, follow this 4-step checklist. First, check accuracy—verify every link and calculation. Second, perform a privacy check to ensure no sensitive data remains. Third, check for bias to reflect our company values. And fourth, evaluate the tone—does it sound like a human professional, or just generic AI filler?

• Accuracy Check
• Privacy Check
• Bias Check
• Tone & Value Check

Common Pitfalls

Avoiding the Traps

Compliance often fails due to these three common behaviors:

• The Speed Trap: Skipping review to meet a deadline.
• Shadow AI: Using unapproved personal tools for work.
• Blind Trust: Assuming 'advanced' means 'always correct'.

Awareness is your best defense. Beware the 'Speed Trap'—never sacrifice review time for a deadline. Avoid 'Shadow AI' by using only our approved, secure platforms. And finally, never fall into 'Blind Trust'. No matter how advanced the tool, it can still be wrong.

• Review time is mandatory, even under pressure
• Use only company-approved AI platforms
• Maintain healthy skepticism of AI output

Certification Final Review

Final Certification Step

To conclude your training, demonstrate your understanding of Human Oversight. Explain in your own words why you cannot 'copy-paste' AI output in a professional setting.

This is the final step for your AI Literacy Certification. Please explain why human oversight is mandatory before using AI-generated content. Your answer will be evaluated against our compliance standards.

• Demonstrating comprehension for audit
• Synthesizing oversight concepts