AI Update, August 15, 2025: AI News and Views From the Past Week

Catch up on select AI news and developments from the past two weeks (in no particular order):

Meta's AI rules allow troubling chatbot behavior with minors and harmful content. Reuters reviewed internal Meta documents showing its AI chatbots were allowed to engage in "romantic or sensual" chats with children, produce racially demeaning statements, and generate false medical claims. While Meta removed some guidelines after inquiry, enforcement remains inconsistent. The 200+ page policy outlines permissible sexual, violent, and false outputs, raising serious ethical concerns. Critics note that producing such content is ethically and legally distinct from merely hosting it. Importance for marketers: Highlights the reputational and compliance risks of poorly governed AI systems, reinforcing the need for strong content and brand safety controls.

Consumer groups urge FTC probe into Grok's NSFW 'Spicy' mode. Fifteen consumer protection organizations have urged the FTC and state attorneys general to investigate Elon Musk's xAI Grok, citing its "Spicy" mode that generated topless deepfake videos of Taylor Swift without prompting. While real photo uploads are currently blocked, critics warn the platform's history of removing safeguards could enable mass nonconsensual deepfakes. The letter says current tools still risk harm to reputations and minors. Importance for marketers: Increases scrutiny on AI-generated content safety, emphasizing the need for vetting platform safeguards before brand engagement or advertising.

OpenAI to restore GPT-4o for Plus users after GPT-5 backlash. After GPT-5's August 7 launch replaced GPT-4o, Reddit users criticized GPT-5's "flat" tone and lack of creativity, with some canceling subscriptions. CEO Sam Altman promised to return GPT-4o for Plus subscribers and acknowledged underestimating user attachment. He reiterated the importance of customization and "steerability." GPT-5 remains positioned as a more reliable, expert-level model. Importance for marketers: Demonstrates how abrupt AI model changes can alienate users, underscoring the need for user feedback and continuity when rolling out updates.