Hi all,
To be fair, I didn't have much time to properly read through the
finalised guidelines before the report went out, so I skipped them. I have now caught up on my reading and to be honest it feels simultaneously overwhelming and underwhelming. A ton of well meant recommendations (as guidelines are non-binding), but it also says that VLOPs are free to come up with other measures to mitigate risks.
The recommendations themselves come in several categories:
- reinforcing internal processes (e.g. having a dedicated team and people who are aware of local contexts),
- risk mitigation measures for electoral processes (e.g. pointing users to official election portals, media literacy initiatives, fact-checking labels, demonetisation, recommender systems),
- mitigation measures linked to generative AI (e.g. watermarks, making an effort to ensure AIs are trained on reliable sources, red-teaming exercises),
- cooperation with national authorities, independent experts and civil society organisations,
- specific recommendations for during the election period (put in place an internal incident response mechanism),
- and for after the election period (conduct a post-election review).
I hope this clarifies your question at least partially!
Cheers,
Dimi