WEBVTT

00:00:00.820 --> 00:00:04.980
Before you use a ChatGPT answer, ask one question.

00:00:05.000 --> 00:00:07.560
What happens if it's wrong?

00:00:07.830 --> 00:00:10.610
This fictional meeting summary says the work is due on

00:00:10.630 --> 00:00:13.430
Friday. The notes never set a date.

00:00:13.450 --> 00:00:16.510
Clear wording doesn't make that promise true.

00:00:21.180 --> 00:00:24.180
Before a real task, think about the people and the

00:00:24.200 --> 00:00:27.120
information involved, and who will read the result.

00:00:27.140 --> 00:00:30.660
A small internal reminder and an employment decision need

00:00:30.680 --> 00:00:33.060
very different levels of review.

00:00:34.130 --> 00:00:37.050
New Zealand privacy law still applies when organisations

00:00:37.070 --> 00:00:39.670
use AI with personal information.

00:00:39.690 --> 00:00:43.650
Keep real staff and client records out of these exercises.

00:00:43.670 --> 00:00:47.210
And never paste passwords or access codes.

00:00:48.240 --> 00:00:50.260
Removing a name isn't enough.

00:00:50.280 --> 00:00:54.400
An address, a unique job title, or a mix of small details

00:00:54.420 --> 00:00:56.400
can still identify someone.

00:00:56.420 --> 00:00:59.100
For practice, invent the whole situation,

00:00:59.120 --> 00:01:01.980
or use the examples in these lessons.

00:01:03.050 --> 00:01:05.090
Then, use four checks.

00:01:05.110 --> 00:01:08.050
Source: compare with the original.

00:01:08.070 --> 00:01:11.730
Facts: verify the important facts and numbers.

00:01:11.750 --> 00:01:15.530
Gaps: find missing details and added promises.

00:01:15.550 --> 00:01:19.050
Impact: decide who must approve the use.

00:01:20.120 --> 00:01:21.980
Open the linked source yourself.

00:01:22.000 --> 00:01:24.300
A link on its own isn't proof.

00:01:24.320 --> 00:01:27.440
If a draft claims a rule, find the original,

00:01:27.460 --> 00:01:30.860
read the passage, and check that it applies to you.

00:01:31.110 --> 00:01:34.950
Record any uncertainty, and ask the right person to review

00:01:34.970 --> 00:01:37.650
decisions with real consequences.

00:01:38.720 --> 00:01:42.580
Next, your team needs a policy people can actually follow.

00:01:42.600 --> 00:01:45.060
The course starter names the approved tool,

00:01:45.080 --> 00:01:47.580
the allowed tasks and data, the reviewer,

00:01:47.600 --> 00:01:49.720
and the incident contact.

00:01:49.990 --> 00:01:52.950
Fill it in for an invented organisation first.

00:01:52.970 --> 00:01:56.290
Then ask your manager and privacy officer to adapt it.

00:01:56.310 --> 00:01:59.430
The template isn't a privacy assessment.

00:01:59.700 --> 00:02:01.860
And say what happens after a mistake:

00:02:01.880 --> 00:02:04.600
who to contact, and how to report it.

00:02:04.620 --> 00:02:08.320
Agree on that before anyone uses real records.

00:02:09.390 --> 00:02:10.753
Now, your next week.

00:02:10.773 --> 00:02:12.070
Five minutes a day.

00:02:12.090 --> 00:02:14.573
Review your settings, draft an email,

00:02:14.593 --> 00:02:17.536
try a quote note, summarise meeting actions,

00:02:17.556 --> 00:02:20.500
check the small dataset, revise a procedure,

00:02:20.520 --> 00:02:23.210
and talk through the policy.

00:02:23.420 --> 00:02:25.920
Keep the source, the prompt, the revision,

00:02:25.940 --> 00:02:27.660
and a quick check note.

00:02:27.680 --> 00:02:31.520
Notice which checks come easily, and which need help.

00:02:32.570 --> 00:02:35.590
Then finish the final quiz, and review your progress.

00:02:35.610 --> 00:02:37.530
You can keep practising for free.

00:02:37.550 --> 00:02:39.850
Share one useful habit with a colleague,

00:02:39.870 --> 00:02:41.470
or your manager.

00:02:41.720 --> 00:02:44.720
If you'd like guided help, the courses page lists more

00:02:44.740 --> 00:02:47.796
courses and workshops, including half-day workshops for

00:02:47.816 --> 00:02:48.980
teams.

00:02:49.250 --> 00:02:52.273
Pick the next step that fits your job and your team.

00:02:52.293 --> 00:02:54.858
This is independent training from The Colab.

00:02:54.878 --> 00:02:57.510
We're not affiliated with OpenAI.
