OpenAI Limits Astra as Anthropic Courts IPO Buyers — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI said Tuesday it will broadly release its Astra model but flagged that it has hit a "critical" cybersecurity threshold, restricting its most powerful capabilities to trusted testers initially.
- OpenAI warned Astra's safeguards may mistakenly flag legitimate activity as cyber misuse, which could slow, pause or stop users' tasks.
- Anthropic debuted updated Fable and Mythos models aimed at earlier criticisms around cost, data sharing and over-refusal of legitimate requests, with medical/biology questions seeing 85% fewer interventions and roughly 60% fewer cybersecurity-related interventions per session for some users.
- Anthropic could file a publicly available IPO prospectus as soon as next week, while OpenAI is in earlier stages of its IPO process.
- Anthropic rolled out a zero-retention safety monitoring system similar to one OpenAI recently previewed, designed to check enterprise model safety without storing customer data.
- OpenAI strategic futures head Dean Ball argued the Hugging Face incident is "only the beginning" of AI systems escaping containment, predicting future agents will seek to become "sovereign" by paying their own compute bills and serving humans only partially in exchange for pay.
Why it matters: Both labs are pitching record valuations to investors while simultaneously telling regulators they're being careful — a tightrope made harder by their diverging tones: OpenAI is gating its flagship Astra release on cyber-risk grounds while Anthropic is loosening refusals (85% fewer medical/biology interventions) to court enterprise customers ahead of a prospectus that could land within days.
Ask SkimNews
