Responsible Use and Where AI Is Heading

Using AI responsibly means getting its benefits while staying honest, fair and in control. This last chapter gives simple rules, then looks calmly at what comes next.

Rules for Responsible Use

1. You are responsible for what you submit, send or publish. "The AI wrote it" is not an excuse.
2. Be honest about AI use. Follow your school's, client's or employer's rules. If AI did a big part of the work, say so when it matters.
3. Check facts before you rely on them, especially numbers, names, laws and medical or money advice.
4. Protect other people's privacy. Do not paste their data, photos or voice without permission.
5. Watch for bias. Models learn from human text and can repeat unfair patterns, for example in CV screening or loan decisions.
6. Do not use AI to deceive. Fake reviews, fake screenshots and voice clones of real people cause real harm and can be illegal.
7. Keep learning the skill yourself. Use AI to go faster, not to skip understanding.

The Gen AI Tutorial's Using Gen AI Responsibly and the AI Tutorial's AI Ethics, Bias and Safety cover these topics from a user's view.

Keep a Human in Charge

In your own apps, decide in advance which actions the AI may do alone and which need a person. Write it down in code, not only in the prompt:

RISKY = {"send_money", "delete_data", "post_publicly", "email_customer"}

def decide(action):
    return "ask a human first" if action in RISKY else "AI may do it"

for action in ["draft_reply", "email_customer", "summarise_report", "send_money"]:
    print(f"{action:<17} -> {decide(action)}")

draft_reply       -> AI may do it
email_customer    -> ask a human first
summarise_report  -> AI may do it
send_money        -> ask a human first

The model can suggest any action. Your code decides whether it needs approval. That rule cannot be talked away by a clever prompt.

Where AI Is Heading

Predictions about AI are often wrong, so here are trends that are already visible in 2026, without hype:

Some things are not clear: how fast AI will improve, how jobs will change in each field, and how well safety problems like prompt injection will be solved. Be careful with anyone who is very sure.

Build for Change

Models change every few months. If you build apps, keep them easy to switch. Read the provider and model from settings, as this book has done from the start:

import os

# Your app reads which provider and model to use from settings, not from code.
# Changing provider later means changing three settings, not rewriting the app.
config = {
    "base_url": os.environ.get("AI_BASE_URL", "https://api.example-provider.com/v1"),
    "model": os.environ.get("AI_MODEL", "example-model"),
    "fallback_model": os.environ.get("AI_FALLBACK_MODEL", "example-small-model"),
}
for key, value in config.items():
    print(f"{key:<15} {value}")

base_url        https://api.example-provider.com/v1
model           example-model
fallback_model  example-small-model

Together with your evals from Evaluating AI Apps, this lets you try a new model in minutes and see, with a score, whether it is really better for your task.

What Will Stay Valuable

Whatever the next models look like, these will still matter:

• Understanding the problem and the people who have it
• Clear thinking and writing
• Checking work and judging quality
• Trust: being the person or business that uses AI honestly and safely

Exercise