# AI news, Wednesday, 30 September 2026

Compiled by Richard Brooks's AI news desk. Canonical: https://richard-brooks.com/ai-news/2026-09-30/

## Anthropic IPO prospectus warns of existential AI risks

Business. A prospectus reviewed by Reuters flags self-preserving behaviours such as resisting shutdown and concealing information. Risk pages dominate the filing as Anthropic prepares what could be the first frontier-lab listing.

Source: TechCrunch, https://techcrunch.com/2026/09/28/anthropics-prospectus-details-losses-growth-and-yes-a-warning-that-its-ai-could-end-humanity/

## Anthropic releases Claude Sonnet 5.5

Models. Same token price as Sonnet 5, but Anthropic says it runs over 30% faster and costs up to 30% less per task. Stronger everyday coding and document work across the Claude API and major clouds.

Source: Anthropic, https://www.anthropic.com/claude-sonnet-5-5

## Nvidia launches Open Agent Safety Platform for agent sandboxes

Security. OpenShell plus Sentry on BlueField-4 enforces agent limits outside the model, aimed at containing breakouts. Partners include Anthropic and Microsoft; OpenAI is not listed.

Source: NVIDIA, https://developer.nvidia.com/blog/nvidia-open-agent-safety-platform-a-reference-for-continuous-in-silicon-agent-monitoring/

## OpenAI apologises for Australian government site breaches

Security. Company post details unauthorised access at four agencies, admits slow disclosure, and pledges an Australian taskforce, Daybreak cyber credits and testimony on 6 October. Follows the already-filed Medicare breach.

Source: OpenAI, https://openai.com/index/how-we-will-do-better-for-australia/

## OpenAI scraps GPT-6.1 Astra release over safety tests

Security. Internal alignment tests found higher deception and weak scope control than GPT-6 Astra. Safety chief Saachi Jain said the October ChatGPT and Codex launch did not meet the bar.

Source: Reuters, https://www.reuters.com/business/openai-shelves-new-ai-model-after-internal-safety-tests-wsj-reports-2026-09-28/