Email immutability matters more in a world with AI (www.fastmail.com)

Fastmail argues that as generative AI makes produced text harder to distinguish from human writing and the web becomes easier to rewrite, email’s immutability becomes increasingly valuable: each sent message is a fixed, personal record you can rely on for provenance and auditing when web pages and other sources are altered. That perspective matters to the AI/ML community because it highlights the need for reliable ground truth, tamper-resistant archives, and tools that surface provenance and timestamps to combat automated rewriting, hallucination, and deliberate revisionism. The piece frames email as an “electronic memory” that preserves accountability in a landscape where lowering the cost of changing history is a real risk. On policy and engineering, Fastmail says it will keep enabling user-driven AI integrations while enforcing privacy and performance safeguards. Staff and vendor use of AI must comply with data-protection rules; any AI-generated content or code requires human review and second-set-of-eyes checks; bias, hallucination and accountability must be actively monitored; and automated decisions must include human appeal paths. Vendors using AI for translation, categorization or abuse detection must uphold customer protections. The takeaway for practitioners: design for auditability and data protection, require human oversight of model outputs, and treat immutable personal records (like email) as critical anchors for provenance in AI systems.