WEBVTT - generated by Videoportal der Uni Paderborn

1
00:00:01.000 --> 00:00:04.800
Welcome to this presentation on “Using AI
in Higher Education – a brief

2
00:00:04.800 --> 00:00:09.000
legal overview.” We will be looking at some
key legal basics

3
00:00:09.000 --> 00:00:14.240
you should be aware of if you want to use generative
AI in your studies. There will also be

4
00:00:14.240 --> 00:00:19.840
a short aside covering the legal aspects of
students recording lectures.

5
00:00:19.840 --> 00:00:25.240
One quick note on the license—this presentation
is published under a CC BY-SA license. This
means

6
00:00:25.240 --> 00:00:29.320
you are free, for example, to share it with
others, (as long as you provide appropriate
credit and share it under the same terms).

7
00:00:30.920 --> 00:00:36.120
But let us start with the use of AI. The use
of generative AI systems in higher education
is

8
00:00:36.120 --> 00:00:41.080
now generally encouraged and well-established
– but teaching staff can decide

9
00:00:41.080 --> 00:00:46.600
whether it is allowed in exams. When using
AI, it is important to comply with

10
00:00:46.600 --> 00:00:52.000
legal requirements and, in this regard, to
respect the rights of others. A lot of things
that are technically easy

11
00:00:52.000 --> 00:00:56.160
to do can still be subject to legal restrictions.
Different areas

12
00:00:56.160 --> 00:01:01.000
of law can be involved here—like data protection,
personality rights—and

13
00:01:01.000 --> 00:01:07.000
in that context also fundamental rights—copyright
law, and the requirements set out in the EU
AI Act.

14
00:01:09.000 --> 00:01:13.000
Let us start with data protection law. It is
always about how personal

15
00:01:13.000 --> 00:01:19.000
data is processed and whether—or to what
extent—that is allowed. But what actually
counts as personal data?

16
00:01:19.000 --> 00:01:23.000
It includes, for example, names, contact details,
student ID numbers, and CVs, but also things
like a computer’s IP address

17
00:01:23.000 --> 00:01:29.000
that can be traced back to the account holder.
It can even

18
00:01:29.000 --> 00:01:32.000
include a handwritten math exam or an oral
exam. That means handwriting or

19
00:01:32.000 --> 00:01:37.000
a person’s voice can already be personal
data, because they might allow you to identify
the student.

20
00:01:38.000 --> 00:01:42.000
When you enter data into an AI system, personal
data is usually

21
00:01:42.000 --> 00:01:47.000
transferred to third countries – most often
the United States. This can include: your prompts,

22
00:01:47.000 --> 00:01:53.000
any files you upload, as well as your IP address,
and so on. These

23
00:01:53.000 --> 00:01:58.000
files – or data – are then processed by
the AI system and may even

24
00:01:58.000 --> 00:02:03.000
be used for training, depending on the settings
you have chosen. It is important to know

25
00:02:03.000 --> 00:02:08.000
that processing personal data always requires
a legal basis: either a

26
00:02:08.000 --> 00:02:13.000
law allows it, or the person has given consent
– but neither is usually the case.

27
00:02:13.000 --> 00:02:17.000
So, as a general rule: Do not upload or enter
any personal data

28
00:02:17.000 --> 00:02:24.000
into AI systems.  Exceptions may apply, however,
to AI systems hosted or installed locally on
your

29
00:02:24.000 --> 00:02:28.000
computer. Examples include Ollama or LM Studio.

30
00:02:28.000 --> 00:02:34.000
From a legal perspective, using such systems
is somewhat comparable

31
00:02:34.000 --> 00:02:39.000
to storing files on your own computer. From
a technical standpoint,

32
00:02:39.000 --> 00:02:43.000
you need to make sure your system has enough
RAM.   Let us now turn to the question of the
extent to

33
00:02:43.000 --> 00:02:49.000
which personal rights and fundamental rights
may be affected. If you use personal data,

34
00:02:49.000 --> 00:02:53.000
photos, and similar content in your prompts,
this may infringe

35
00:02:53.000 --> 00:02:56.000
on a person’s right to control their own
images and information. In such cases, there
may be violations of

36
00:02:56.000 --> 00:03:00.000
general personality rights, the right to the
protection of personal data,

37
00:03:00.000 --> 00:03:06.000
and the right to one’s own image, which is
considered part of general personality rights.

38
00:03:06.000 --> 00:03:10.000
As a general principle: Those affected have
the right to decide for themselves how their
data,

39
00:03:10.000 --> 00:03:15.000
photos, videos, et cetera are used. The problem
is: once data is in the system,

40
00:03:15.000 --> 00:03:20.000
it is very difficult to remove it again.

41
00:03:20.000 --> 00:03:25.000
So, keep in mind: your inputs can violate these
rights – as a rule of thumb, you should

42
00:03:25.000 --> 00:03:30.000
avoid this –at the same time, be aware that
there can be exceptions,

43
00:03:30.000 --> 00:03:35.000
for example when the information about a person
is already publicly accessible – for instance
the authorship

44
00:03:35.000 --> 00:03:40.000
of an article.   Now let us move on to the

45
00:03:40.000 --> 00:03:44.000
copyright perspective. This is interesting
in two respects: first, in terms of

46
00:03:44.000 --> 00:03:48.000
the input, and second, in terms of the output.
Let us start with the input –  that is,

47
00:03:48.000 --> 00:03:53.000
what you enter into the system, your prompts.
Working with your own drafts and materials

48
00:03:53.000 --> 00:03:57.000
is generally not a problem. However, the situation
can be different if you want to upload someone
else’s

49
00:03:57.000 --> 00:04:02.000
protected work. Academic papers, presentations,

50
00:04:02.000 --> 00:04:07.000
images, et cetera are usually protected by
copyright. Most AI systems

51
00:04:07.000 --> 00:04:11.000
also require, in their terms of use, that you
actually have the rights to whatever you upload.

52
00:04:11.000 --> 00:04:16.000
And more generally, you need copyright permission
for what

53
00:04:16.000 --> 00:04:21.000
you are doing—if it is relevant under copyright
law. There is no

54
00:04:21.000 --> 00:04:24.000
clear legal precedent for this issue yet, but
the prevailing view is that

55
00:04:24.000 --> 00:04:29.000
uploading content to AI systems is likely to
be considered copyright-relevant – specifically,
as a form of reproduction –

56
00:04:29.000 --> 00:04:34.000
so you would need permission.  At this point,
the key is to avoid copyright

57
00:04:34.000 --> 00:04:38.000
infringements and to use material that is either
no longer

58
00:04:38.000 --> 00:04:43.000
protected or licensed in a way that allows
you to use it—such as under a Creative Commons
license, for example CC BY

59
00:04:43.000 --> 00:04:49.000
or CC BY-SA. In those cases, you should generally
be fine.  Now let us move on to the output
side. To keep

60
00:04:49.000 --> 00:04:53.000
it short: in most cases, AI-generated output
is not protected by copyright. There is already

61
00:04:53.000 --> 00:04:58.000
an initial court decision on this from the
Amtsgericht München (Munich Local Court).
In that case,

62
00:04:58.000 --> 00:05:02.000
someone had created graphics using AI and multiple
prompts, but the court still

63
00:05:02.000 --> 00:05:08.000
found that the user’s personal creative contribution
was not sufficient, so no

64
00:05:08.000 --> 00:05:14.000
copyright protection applied. The reason is
that the AI system itself cannot be considered
an author, because it is

65
00:05:14.000 --> 00:05:20.000
not a natural person. As a result, the content
is effectively in the public domain—that
is,

66
00:05:20.000 --> 00:05:25.000
it can be freely used by others. However—and
this actually comes up quite often—

67
00:05:25.000 --> 00:05:30.000
AI-generated output can be further edited.
And that additional

68
00:05:30.000 --> 00:05:36.000
editing can lead to copyright protection, even
if the original material was AI-generated.

69
00:05:36.000 --> 00:05:40.000
In that case, it becomes important who made
those changes.

70
00:05:40.000 --> 00:05:44.000
If copyright protection arises—meaning the
work reaches the required level of originality—you
would need

71
00:05:44.000 --> 00:05:51.000
to ask that person for permission. You also
need to be careful if the output contains recognizable
reproductions of existing

72
00:05:51.000 --> 00:05:57.000
copyrighted works. In that case, you should
not use it further—or you

73
00:05:57.000 --> 00:06:03.000
should seek permission if you plan to do so.
The copyright assessment we just discussed

74
00:06:03.000 --> 00:06:08.000
also affects whether you can license AI output
under Creative Commons.

75
00:06:08.000 --> 00:06:14.000
That is because the possibility of using a
CC license depends on the type of material
you are

76
00:06:14.000 --> 00:06:20.000
dealing with. Are we talking about purely AI-generated
content?

77
00:06:20.000 --> 00:06:25.000
Or AI-generated content that you have further
edited? Or your own protected work that

78
00:06:25.000 --> 00:06:30.000
you have had processed by an AI system? One
key point to keep in mind is: Creative Commons
licensing only

79
00:06:30.000 --> 00:06:35.000
really makes sense if the material reaches
the required level of originality—otherwise,
there are no rights to license in the first
place.

80
00:06:35.000 --> 00:06:40.000
So, you always need to check whether the material
meets that threshold. If you are dealing

81
00:06:40.000 --> 00:06:46.000
with purely AI-generated content—so there
is no sufficient originality—you do not

82
00:06:46.000 --> 00:06:51.000
need a CC license. In that case, simply labeling
it as “public domain” is enough.

83
00:06:52.000 --> 00:06:56.000
If your own edits add enough originality, then
you can

84
00:06:56.000 --> 00:07:02.000
choose a Creative Commons license as you wish.
And if your original material was

85
00:07:02.000 --> 00:07:06.000
already protected—so it had sufficient originality—and
the

86
00:07:06.000 --> 00:07:13.000
AI’s contribution was only minor, you can
also apply a CC license of your choice.

87
00:07:13.000 --> 00:07:18.000
Let us now turn to labeling requirements for
AI-generated content.  As a general rule,

88
00:07:18.000 --> 00:07:24.000
transparency is helpful and advisable in most
cases—but

89
00:07:24.000 --> 00:07:30.000
at the moment, there is no general obligation
to label AI-generated content. However, academic
work

90
00:07:30.000 --> 00:07:36.000
is subject to the principles of good scientific
practice. According to Guideline 14 of the
German Research Foundation (DFG)

91
00:07:36.000 --> 00:07:42.000
guidelines on good academic practice, an author
is someone

92
00:07:42.000 --> 00:07:46.000
who has made a genuine and traceable contribution
to the content of a scientific work.

93
00:07:47.000 --> 00:07:51.000
So in general, it is important to be transparent
about how you worked and how

94
00:07:51.000 --> 00:07:57.000
you arrived at your results. And quite often,
there are also specific requirements

95
00:07:57.000 --> 00:08:02.000
set by examiners or included in declarations
of independent work.

96
00:08:02.000 --> 00:08:07.000
Such requirements can also arise from university
regulations,

97
00:08:07.000 --> 00:08:11.000
examination rules, or AI policies. So, if there
are specific rules,

98
00:08:11.000 --> 00:08:15.000
you definitely need to follow them—and make
your use

99
00:08:15.000 --> 00:08:22.000
of AI transparent accordingly. In many cases,
examiners will also specify how exactly this
should be done.

100
00:08:22.000 --> 00:08:28.000
For the sake of completeness, there is also
a legal obligation coming into force from August
2026. It applies to so-called

101
00:08:28.000 --> 00:08:34.000
“deepfakes”—that is, deliberately deceptive
AI-generated content—and

102
00:08:34.000 --> 00:08:39.000
to AI-generated information on matters of public
interest. These are

103
00:08:39.000 --> 00:08:46.000
usually not the main concern in student work,
but they are still worth mentioning here.

104
00:08:46.000 --> 00:08:50.000
Now let us move on to the second topic I mentioned:
unauthorised audio

105
00:08:50.000 --> 00:08:55.000
or video recordings of lectures by students.
It does happen that students

106
00:08:55.000 --> 00:08:59.000
want to record lectures—and not infrequently,
this is done

107
00:08:59.000 --> 00:09:04.000
without permission. Sometimes, these recordings
are even shared or published.

108
00:09:04.000 --> 00:09:08.000
The reasons can vary. For example, students
may want to review the material at

109
00:09:08.000 --> 00:09:14.000
their own pace, or they may be struggling to
understand certain parts. However, many people
are not aware

110
00:09:14.000 --> 00:09:18.000
that making such recordings can affect or even
violate a number of rights—fundamental rights,

111
00:09:18.000 --> 00:09:23.000
personality rights, copyright, data protection
law, and university regulations. In some cases,

112
00:09:23.000 --> 00:09:27.000
this can even have criminal implications. We
will take a closer look at

113
00:09:27.000 --> 00:09:33.000
which rights may be affected on the following
slides. Let us start with personality rights,

114
00:09:33.000 --> 00:09:38.000
or more broadly, fundamental rights. The general
right of personality, (which includes things
like privacy, control over one’s image, and
personal data),

115
00:09:38.000 --> 00:09:44.000
can be affected in several ways: the right
to self-determination and self-presentation,

116
00:09:44.000 --> 00:09:48.000
the right to one’s own words and voice, the
right to the protection of personal

117
00:09:48.000 --> 00:09:53.000
data, and the right to control the use of one’s
own image.

118
00:09:53.000 --> 00:09:58.000
It is important to keep in mind that this

119
00:09:58.000 --> 00:10:02.000
can affect both lecturers and students—especially
those who actively

120
00:10:02.000 --> 00:10:07.000
in class. But perhaps even more importantly,
both lecturers and

121
00:10:07.000 --> 00:10:11.000
students should be able to express themselves
freely, without having to worry about the public
impact of their words,

122
00:10:11.000 --> 00:10:17.000
their voice, or their appearance. Let us now
look at the relevance from the

123
00:10:17.000 --> 00:10:22.000
perspective of data protection law, university
law, and also house rules. As we have already
discussed, the processing

124
00:10:22.000 --> 00:10:26.000
of personal data—that is, recording, storing,
and sharing it—is

125
00:10:26.000 --> 00:10:31.000
only allowed if there is a legal basis. But
in this case, there is generally no such legal
basis—neither under

126
00:10:31.000 --> 00:10:37.000
the GDPR (DSGVO) nor under university law.
In addition, house rules—provided for in
the Higher Education Act of North Rhine-Westphalia
(NRW)—

127
00:10:37.000 --> 00:10:42.000
give the authority to decide whether recordings
are allowed in certain spaces.

128
00:10:42.000 --> 00:10:48.000
In practice, this authority is exercised by
the respective lecturers. So, the key point
is to avoid violations and to be

129
00:10:48.000 --> 00:10:53.000
aware of the value of an open and trusting
learning environment. Because if people

130
00:10:53.000 --> 00:10:56.000
feel that they are being recorded—or might
be recorded—valuable discussion and exchange,

131
00:10:56.000 --> 00:11:01.000
both with lecturers and among students, can
be lost.

132
00:11:02.000 --> 00:11:07.000
Now let us turn to copyright. Recordings can
also affect copyright,

133
00:11:07.000 --> 00:11:11.000
because the way teaching content is formulated
or presented—for example, longer passages
of text,

134
00:11:11.000 --> 00:11:16.000
full texts, or presentation materials like
photos, diagrams, or graphics—

135
00:11:16.000 --> 00:11:21.000
is often protected by copyright. Even the way
in which content is explained

136
00:11:21.000 --> 00:11:25.000
or discussed in a lecture or presentation can

137
00:11:25.000 --> 00:11:30.000
be protected. So, when recordings are made,
this creates reproductions of the teaching
and

138
00:11:30.000 --> 00:11:36.000
presentation materials—and that generally
requires permission, which is usually not given.

139
00:11:36.000 --> 00:11:40.000
This becomes even more relevant from a copyright
perspective if recordings are shared,

140
00:11:40.000 --> 00:11:46.000
posted on social media, or entered into generative
AI systems. So this

141
00:11:46.000 --> 00:11:52.000
is something you should refrain from doing
under any circumstances. Finally, a few words
on criminal law.

142
00:11:52.000 --> 00:11:57.000
In more serious cases, there can actually be
criminal consequences. Section 201

143
00:11:57.000 --> 00:12:03.000
of the German Criminal Code (StGB) protects,
among other things, the confidentiality of
the spoken word. It makes it a

144
00:12:03.000 --> 00:12:09.000
criminal offense to record or use speech that
is not intended to be public. So, when it comes

145
00:12:09.000 --> 00:12:13.000
to recording non-public lectures, criminal
liability under this provision cannot be ruled
out—

146
00:12:13.000 --> 00:12:17.000
especially if recordings are made secretly
and then

147
00:12:17.000 --> 00:12:22.000
shared on social media. So in summary: do not
make recordings—and definitely do not

148
00:12:22.000 --> 00:12:27.000
share them online. If you have ever thought
about recording a lecture,

149
00:12:27.000 --> 00:12:33.000
ask yourself how you or others would feel about
being recorded without consent—and instead,

150
00:12:33.000 --> 00:12:38.000
try to talk to your lecturer. We have now covered
a lot of important points—

151
00:12:38.000 --> 00:12:42.000
so thanks for sticking with me. Let us briefly
sum up the key takeaways:

152
00:12:42.000 --> 00:12:47.000
Treat other people’s data and their work
with respect when using AI systems.

153
00:12:47.000 --> 00:12:52.000
Purely AI-generated content can generally be
used freely—but if you further edit that
output,

154
00:12:52.000 --> 00:12:57.000
copyright protection may arise. If the AI output
contains third-party protected works,

155
00:12:57.000 --> 00:13:03.000
you need to be careful. Using a Creative Commons
license for AI output only makes sense if

156
00:13:03.000 --> 00:13:09.000
the material meets the required level of originality.
Refrain from recording lectures—and instead,

157
00:13:09.000 --> 00:13:16.000
try to engage with your lecturers to discuss
how things can be improved.