Inaccurate automated translations on Meta’s social media platforms
8.11.2023
Question for written answer E-003296/2023
to the Commission
Rule 138
Samira Rafaela (Renew)
Following up on Written Question E-003111/2023[1] regarding shadow-banning on major social media platforms, I would now like to make the Commission aware of recent, and deeply worrying, occurrences of ‘inaccurate automated translations’.
Meta has officially apologised after falsely inserting the word ‘terrorist’ into the profile biographies of Palestinian Instagram users[2]. Concern about these obvious, often one-sided, digital biases creeping into people’s daily use of social media is growing by the day.
- 1.Could the Commission elaborate on the specific mechanisms or audits it plans to implement in order to thoroughly investigate potential freedom of speech violations by platforms like Meta?
- 2.If these incidents are found to violate freedom of speech and contravene the rules outlined in the Digital Services Act, what steps will the Commission, as the EU’s executive authority, take to enforce penalties on these companies?
- 3.In cases of inaccurate automated translations or similar errors, what steps will the Commission take to gather comprehensive data and insights in order to understand the root causes of the errors? For example, on whether they result from a misconfiguration within the automated translation system, or were the consequence of a direct, misguided human command.
Submitted: 8.11.2023
Last updated: 21 November 2023