AI safety and judgment, part 2: hallucinations, bias, copyright and policy
Claude can be confidently wrong, quietly unfair, or borrow someone else's work. Know the four risks and the one-line disclosure habit.
Hallucinations: confident and wrong
A hallucination is an answer that sounds right and is not. Claude might invent a statistic, a quote, a legal requirement, a product feature or a person. It does not do this on purpose and it does not warn you. The sentence reads exactly like a true one.
Where hallucinations happen most:
- Specifics with no source: figures, dates, names, citations, phone numbers.
- Things that should exist: “the county permit form”, “the standard clause”. Claude may describe one that does not exist.
- The edges of a document: when asked to summarise, it may fill a gap with a plausible guess.
The defence is the two-minute check from lesson 6 plus one rule: a fact that matters needs a source you have opened. If Claude cannot point at where a claim came from, treat it as unverified.
The three hotspots and the defence, together:
Bias: quietly unfair
Claude learned from human writing, and human writing carries assumptions. Left unchecked, AI drafts can:
- Assume a nurse is “she” and an engineer is “he”.
- Rank job applicants in ways that track names, schools or gaps rather than skills.
- Write to one customer more warmly than another based on a name or neighbourhood.
Two habits help. First, never let an AI make the decision about a person: hiring, firing, pricing, credit, housing. Use it to draft and organise, then decide yourself. Second, when you ask for anything that compares people, ask Claude to list the criteria it used and check them.
Copyright: whose words are these?
Claude’s output is generally yours to use at work, but two cautions:
- Do not ask it to reproduce someone else’s work. “Rewrite this competitor’s brochure so it’s ours” is a copyright problem no matter who types it.
- Images and logos of other companies, characters or people are off limits for your marketing. Make your own.
If Claude produces a long passage that reads like a quote from a real book, article or song, assume it may be one and do not publish it.
Your company AI policy
Most small businesses now have a short AI policy, often one page. It usually says:
- Which tools and accounts are approved.
- What may be pasted (the three buckets from part 1).
- Which tasks need a human to decide.
- When AI use must be disclosed.
If you have not read yours, ask for it today. If your company does not have one, the rules in these two lessons are a safe default until it does.
Disclosure: the one-line habit
Sometimes people need to know that AI helped. Rules of thumb:
| Situation | Disclose? | How |
|---|---|---|
| Internal draft you edited and own | Usually no | Nothing needed; it is your work |
| Customer-facing text you edited and checked | Follow policy | Often no, as long as a person reviewed it |
| Anything a customer might assume a human did personally, such as a condolence note or a personal reference | Yes, or write it yourself | “I used an AI tool to help draft this” |
| Formal documents: legal, medical, financial advice | Yes, and a qualified person must review | Note in the document or cover email |
| Images or audio that look real | Yes, always | Caption or label |
When unsure, disclose. Nobody has ever been angry to learn you were honest.
The same rules as a quick decision path:
Try it now
Find your company AI policy, or ask who owns it. Read it against lessons 7 and 8 and write down one thing it adds and one thing it does not cover. If there is no policy, write three sentences that could become one and send them to your manager.
Remember: a fact that matters needs a source you opened, people decisions stay with people, do not borrow others’ work, and disclose when a reader would want to know.
Checking your progress…