GE BSI updated publication on opportunities and risks of large AI language models.
Follows,GE BSI May 2023 paper on AI language opportunities/risks, see #172175.
Guide primarily aimed at companies and authorities wanting to integrate large language models (LLMs) into workflows and can be used as basis for risk analysis.
Update
As part of the update, document has been restructured with risks divided into risks due to proper use, risks due to improper use, risks resulting from attacks on the LLM.
Further countermeasures also identified and linked to corresponding risks in a table.
The restructuring created the possibility of integrating text-generating AI models into the publication in the future, as well as those that process and output other media.