OpenAI Astra Risk Meets Free ChatGPT Push

OpenAI is tightening controls around a cyber-capable model while widening access to its consumer AI.

🔐 Astra Forces a Higher Cybersecurity Bar

Decoded: OpenAI said Aug. 7 that internal evaluations of its upcoming Astra model showed enough agentic coding and cybersecurity progress that it cannot rule out Critical capability under its Preparedness Framework. OpenAI defines that threshold as independently finding functional zero-day exploits across many hardened critical systems or executing novel end-to-end attacks from a high-level goal. The company paused Astra work that does not meet strengthened controls and added isolated testing, restricted network and tool access, model-weight protections, sandboxing and universal monitoring. Source: OpenAI and Reuters

Why it matters: Advanced cyber capability is becoming a release-management and infrastructure cost, not just a benchmark. OpenAI's controls set a reference point that enterprise buyers may also demand from Microsoft, Google and Meta before agents receive broader permissions.

🤖 ChatGPT Expands the Free-Tier Funnel

Decoded: OpenAI said Aug. 6 that GPT-5.6 Luna will become the default for Free and Go users, with unlimited text chats and a Think button arriving next week, subject to abuse safeguards. It also updated GPT-5.6 Sol for Plus and Pro users with a reasoning-effort slider. In OpenAI's internal evaluation of financial, medical and legal prompts, answers containing at least one factual error were about 62% less common with Luna and 68% less common with Sol than with GPT-5.5 Instant. The Work and Codex version of Sol is unchanged. Source: OpenAI

Why it matters: Unlimited text access can expand ChatGPT's user funnel, but it also shifts attention to inference cost and paid conversion. The product split lets OpenAI widen consumer distribution without changing the model serving Microsoft-linked enterprise and developer workloads.

Stay decoded. See you tomorrow.

— The Get AI Decoded Team