In a move reflective of the growing concern around the use of artificial intelligence (AI) to spread misinformation, Midjourney, an AI image generator platform, has implemented a ban on image prompts involving both Donald Trump, the presidential hopeful, and Joe Biden, the current president. This decision arrives amidst warnings from experts and advocates about the potential misuse of AI technology to influence voter decisions and disseminate false information in the lead-up to the 2024 presidential election.
Key Highlights
- Midjourney bans prompts including Donald Trump and Joe Biden.
- The ban aims to prevent the spread of election misinformation.
- Attempts to generate such images trigger a “Banned Prompt Detected” warning.
- Midjourney CEO emphasizes the difficulty of moderating political speech.
- Similar actions taken by other AI platforms to combat misinformation.
- Midjourney criticized for previously lacking stringent policies against election disinformation.
Midjourney’s recent policy update signifies a significant shift in its approach to moderating content, particularly with political undertones. The platform, previously criticized for its relaxed stance on misinformation, now joins the ranks of other generative AI tools that have introduced restrictions to combat the proliferation of harmful content.
The Decision’s Background and Rationale
The Associated Press reported that tests of Midjourney’s new policy yielded warnings for image prompts depicting politically charged scenarios, such as “Trump and Biden shaking hands at the beach.” Further attempts led to warnings about triggering an abuse alert. Midjourney’s CEO, David Holz, expressed at a March 13 press gathering that while political speech was not the platform’s focus, the challenge of moderating such content necessitated stricter measures.
This move by Midjourney is part of a broader trend among AI platforms to address concerns over their potential to spread misinformation. For instance, Microsoft’s Bing Image Generator and OpenAI’s DALL-E 3 have introduced similar bans targeting content that could be misused to mislead or harm the public.
Industry-Wide Efforts to Curb Misinformation
The landscape of generative AI is increasingly coming under scrutiny for how it handles content that could contribute to misinformation, especially in the sensitive context of elections. OpenAI, for instance, has revamped its DALL-E 3 with a multi-tiered safety system designed to limit the generation of content that is violent, hateful, or otherwise inappropriate. Specific guidelines have been issued to prevent election disinformation, including declining requests for images of real people, including political candidates.
Challenges and Criticisms
Midjourney’s approach to tackling election misinformation has not been without its challenges. Prior to this ban, the platform was one of the few major AI image generators that had not adopted a formal stance or policy against election disinformation. It also did not sign a voluntary industry pact aimed at adopting precautions against deepfakes and disinformation.
A report by the Center for Countering Digital Hate highlighted Midjourney’s previous shortcomings in moderating content, showing that it lagged behind other tools in preventing election disinformation, with a 65 percent failure rate in identifying such content.
Looking Ahead
Midjourney’s decision to ban image prompts of Joe Biden and Donald Trump is a clear acknowledgment of the role AI technology plays in the digital ecosystem and its potential impact on democratic processes. As AI continues to evolve, the challenge for platforms like Midjourney lies in balancing the innovative potential of AI with the imperative to safeguard against its misuse. This policy change marks a critical step in the right direction, reflecting a commitment to responsible AI use amidst the digital age’s complexities.
