‘Retroactive Consent’: Social Media Platforms Make It Hard to Opt Out of Training Their AI Models

Welcome to this Cloud Wars Agent and Copilot Minute. In these discussions, I’ll be analyzing opportunities, impact, and outcomes possible with AI; this episode addresses consent and AI training on social media platforms.

Highlights

00:18 — Tech blog Lifehacker conducted an audit across 11 major platforms to determine if users can opt out of AI training on their content. Twitch is highlighted as one of the more consumer-friendly platforms for allowing streamers to opt out of AI training for their videos, clips, and chat logs. The audit reveals that almost every platform uses user content to train AI models, but few make it easy to find or use the opt-out feature.

01:26 — Only four platforms (LinkedIn, Snapchat, Twitch, and X) offer a real first-party toggle for users to opt out of AI training. YouTube allows creators to block third-party companies from training on their channel content but does not offer an opt-out for Google’s own model training. Facebook, Instagram, Threads, and TikTok do not offer meaningful first-party opt-out options unless users live in the EU, where GDPR forces Meta to honor an objection process.

02:12 — None of the platforms offer an opt-in model; instead, they treat existing account activity as retroactive consent. Users must actively seek out the opt-out option, which is often buried deep in settings and set by default to “on.” Most enterprises also struggling with this issue internally: Plenty of companies can name sanctioned AI tools but struggle to track what data is being used by various SaaS vendors and consumer AI tools.

03:55 — Here’s the enterprise version of the problem: vendor contracts often do not reflect updated AI terms. I recommend that organizations inventory their AI vendor policies and evaluate vendors based on their default data use settings.

More of My AI Insights:

• AI Success Requires Integration With Strategy, Apps, and Cloud Infrastructure
• ‘Context Engineers’ Bridge Gap Between AI Agents in Demos, Production
• As AI Matures, Companies Are Building ‘Operating Systems for the Enterprise’
• Anthropic Government Contract, Canadian School Shooting Highlight AI Data Privacy Tensions