230830-aleph-alphaposition-paper-table-ai-act-art-28-28b-final
Dieses Dokument ist Teil der Anfrage „Stellungnahmen von Aleph Alpha an das BMWK“
ALEPH ALPHA: POSITION PAPER TABLE FOR ART. 28 – 28B EU ACT 2023
AI-Act Text proposed by Text adopted by the European Modifications suggested by Why?
Article the EU Parliament (EP) Aleph Alpha (in red)
Commission on 14 June 2023
on 21 April 2021 (P9_TA(2023)0236)
(2021/0106 Info: new text by EP is in bold and italic
(COD)
case, there is a need to train the foundation
model with all possible biases.
There is a risk of a considerable
infringement of rights if the provider of the
foundation model has to predefine "flying
blind" and without knowledge of the use
case, which contents may and may not
occur.
The output is not controlled by the
foundation model provider, but by the
downstream provider, that further
manipulates the system, trains with its own
data and is the only person within the value
chain that can ensure the exact usage
determination of the system, that the legal
situation needs to be measured by.
c) without prejudice to Union or c) without prejudice to Union or
national or Union legislation on national or Union legislation on
Modifications necessary:
copyright, document and make copyright, design and implement an
IP issues are already conclusively
publicly available a sufficiently adequate management of copyright
regulated in Art. 3 and 4 Digital Single
detailed summary of the use of law aspects in relation to text and
Market Directive (EU) 2019/790 and
training data protected under data mining of training data, in
national transposing legislation. This
copyright law. particular with respect to Art. 3 and
amendment underlines the coherence with
4 of Directive (EU) 2019/790
existing intellectual property law.
document and make publicly
available a sufficiently detailed
summary of the use of training data
protected under copyright law.
Page 17 of 17 File: Aleph-Alpha_Position-Paper-Table_AI-Act-Art-28-28b / Aleph Alpha Legal legal@aleph-alpha.com / 16 Aug 2023