Online platforms subject to the EU’s Digital Services Act will be asked to participate in “stress tests” of their content moderation systems before the regulation takes effect, the bloc’s industry commissioner said today.
Thierry Breton made the remarks to journalists today at the EU-US Trade and Technology Council, a trans-Atlantic tech policy summit
being held in Sweden.
The “stress tests” will be voluntary and involve looking at platforms’ procedures for dealing with the spread of disinformation, MLex understands.
"Very large online platforms" including Google, YouTube, Twitter, TikTok, Facebook and Instagram will be required by the DSA to show regulators their systems appropriately manage the risk of their services being used to spread disinformation. The rules will become binding on Aug. 25.
But Breton wants to conduct “stress tests” of these systems before then, to ensure platforms are already in a position to comply when the DSA begins to bite.
Further details about the “stress tests” are expected in June, MLex understands.
In November last year, Breton and Twitter boss Elon Musk agreed to carry out a DSA “stress test” of the platform’s systems.
But Twitter recently signaled its intention to withdraw from the EU's voluntary code of practice on disinformation, which under the DSA will become one of the ways by which platforms can demonstrate their compliance with the binding risk-management rules (see here).
— Comply early with AI Act —
The commissioner is also pushing for early compliance with the forthcoming Artificial Intelligence Act, which is not yet finalized and likely won’t take effect until 2026. Negotiations between national governments and the European Parliament to determine the final version of the law are expected to begin soon.
But once the final version has been passed into law, there will be a two-year countdown before it becomes binding.
On May 24, Breton and Google boss Sundar Pichai agreed to set up a voluntary “AI Pact” to establish temporary standards for firms to follow until the AI Act comes into force (see here).
“We cannot afford to sit back and wait until the regulation becomes applicable (most likely in 2026)," Breton said on Monday in a written media statement. "I will work together with AI developers, EU and non-EU based, to assist them in their preparation and in already implementing the AI Act on a voluntary basis ahead of the legal deadline — that is the idea of the AI Pact.”
