OpenAI's Private Safety Processing takes on Anthropic — SkimNews
SkimNews Take
Reframing competition around data retention rather than model capability shifts enterprise procurement toward auditable data lifecycles—a structural edge in regulated sectors, though the cross-conversation misuse detection itself becomes the next privacy claim to scrutinize.
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI previewed Private Safety Processing, an automated system that monitors for AI misuse across multiple sessions while retaining no customer data — extending its existing Zero Data Retention approach with cross-conversation analysis
- The system sends a "narrowly defined signal" to OpenAI when triggered, and any further data sharing with the company happens at the customer's discretion
- Anthropic announced in July a 30-day retention policy covering user sessions and conversations for "covered models" including Mythos-class models, a move that aggravated enterprise customers handling sensitive data
- Anthropic permits human review of customer data only through a "controlled access path" with a "small set of approved reviewers" and tamper-proof logs
- Competitive backdrop: Anthropic's annualized revenue run rate is reportedly $65 billion with a potential $2 trillion IPO, while OpenAI's Q2 reportedly grew more slowly than its rival
Why it matters: With Anthropic at a reported $65 billion annualized revenue run rate and OpenAI's Q2 reportedly growing more slowly, OpenAI's no-retention safety tool gives enterprise buyers a concrete reason to pick one lab over the other on data-handling grounds — exactly the kind of differentiation that matters most in regulated industries where data residency is a procurement gate.
Ask SkimNews
