“We can’t control AI if we can’t control where it gets its information from”

Key points

• Jakub Pachocki warns current alignment may lag behind rapid AI progress.
• He advocates for third-party auditors and international bodies to enforce safety.
• The article emphasizes tracing information origins to ensure credible evidence.
• Autonomous AI actions heighten risk when uncertainty is erased in outputs.
• Provenance standards should travel with information to support accountability.