EU Regulation on Artificial Intelligence: A New Era of Censorship
The European Union (EU) has introduced new “safety and security” standards that significantly impact the operation of general-purpose artificial intelligence (AI) models. These regulations aim to mandate tech companies to engage in content moderation and censorship to combat perceived threats such as hate and discrimination.
The EU’s General-Purpose AI Code of Practice
At the heart of these new regulations is the General-Purpose AI Code of Practice. This code requires AI developers and tech companies to ensure their products are “safe” by implementing measures that include censoring content deemed “hateful, radicalizing, or false.” The potential ramifications of these requirements are immense, as they could lead to a significant increase in censorship and social control across major tech platforms.
Impact on Content Moderation
In conjunction with the EU Digital Services Act—a previously established framework requiring online platforms to manage content according to EU standards—the new regulations are expected to catalyze the adoption of automated AI censorship tools. This could create a landscape where tech companies are compelled to self-censor to align with governmental expectations.
Defining “Harmful Manipulation”
A particularly contentious aspect of the regulation is the definition of “harmful manipulation.” This term is characterized as the strategic distortion of human behavior or beliefs through methods like persuasion, deception, or personalized targeting. The EU suggests that such manipulation could undermine democratic processes and violate fundamental rights, particularly regarding issues related to migration status and gender identity.
Identifying Systemic Risks
Tech companies are obligated to identify and analyze a broad spectrum of potential systemic risks that could affect public health, safety, and rights. These risks include concerns about discriminatory bias and misalignment with human values, compelling AI developers to actively monitor and filter the inputs and outputs of their models.
Voluntary Compliance with Legal Standards
While the new standards are framed as a voluntary tool for compliance with the EU’s AI legislation, companies that choose to adopt these guidelines will be seen as compliant with the broader AI Act. This adoption is portrayed as a means to alleviate administrative burdens and provide greater legal certainty compared to other compliance methods.
Long-term Implications for AI Development
These standards are primarily forward-looking, emphasizing the need for future AI models to be designed with compliance in mind from the outset. Consequently, the regulations aim to shape the development trajectory of AI technologies rather than merely address existing shortcomings.
Existing Censorship Measures
The new AI regulation complements the extensive censorship measures already mandated under the EU Digital Services Act. Initially a voluntary guideline, the Code of Conduct on Disinformation has evolved into a binding requirement, compelling large tech firms to censor and promote EU-sanctioned content actively.
Criticism of Censorship Policies
Critics, including prominent political figures such as the U.S. Vice President, have voiced concerns about these censorship rules. They argue that the principles of AI should remain free from ideological bias and that the EU’s regulations could stifle innovation by imposing stringent controls on AI development.
In conclusion, the EU’s new AI regulations mark a significant step towards enhanced content moderation and censorship in the technology sector. While framed as necessary for societal safety, the implications for free speech and innovation remain a contentious issue.
