{{first_name | Hey}}, welcome back.
In today’s issue, Anthropic’s pushing for tougher AI safety rules, Oura’s planning $16B IPO, and OpenAI’s GPT-6 Astra is raising the stakes for AI capability and oversight.
These are today's updates.
Simplify SOC 2, ISO 27001, and security compliance with Vanta*
Anthropic backs tougher AI safety rules
Oura's revenue nearly doubles
OpenAI calls GPT-6 Astra AGI
Tools, jobs, resources, and the last issue’s winning prompt ⬇️
Together with Vanta
Your prospects won’t sign without proof of security. A proof of compliance request stalls deals and pulls engineers into audit prep. Vanta gets you compliant fast (SOC 2, ISO 27001, HIPAA+) and keeps you that way.
Get audit-ready without the manual busywork
Respond to security reviews faster
Keep compliance continuously on track
Trusted by 16,000+ companies, including Ramp, Harvey, and Writer
Top News
Anthropic is backing a Massachusetts Senate proposal that would require major AI labs to hire independent evaluators to assess catastrophic risks in their models every 120 days. These are the toughest state-level AI safety rules proposed so far. OpenAI and Google are pushing back, both preferring a lighter model based on Illinois's law, which only requires a yearly self-audit.
Anthropic's head of state policy, Cesar Fernandez, says that the industry shouldn't grade its own homework, while OpenAI's Donnie Fowler argues that a patchwork of different state rules doesn't make it safer; it just creates more confusion.
It's a repeat of the pattern from California's SB 53 last year, where Anthropic was the only major lab to back stricter transparency rules while rivals lobbied against them.
The smart ring maker's SEC filing shows revenue jumped from $697 million to $1.2 billion over the nine months ending in June, compared to the same period a year earlier. Oura has sold 3.6 million rings and has about 5 million paying subscribers, with 85% of members still subscribed a year after signing up.
The company is reportedly seeking a $16 billion valuation, up from $11 billion last October, and hoping to raise $3 billion in the offering. Oura is leaning into its AI angle in the filing, saying its dataset of nearly 42 billion hours of biometric data powers the machine learning models behind its health predictions.
OpenAI's new flagship model, GPT-6 Astra, topped every frontier model on benchmarks spanning computer use, coding, and cybersecurity, according to the company. President Greg Brockman called it the arrival of AGI on a call with reporters. Astra is rolling out now to a limited set of organizations, with wider access for ChatGPT Plus, Pro, Business, and Enterprise users coming over the following days.
The model's cybersecurity score is the part drawing the most attention: Astra hit 100% on ExploitBench, and OpenAI says it can find and chain together unknown vulnerabilities without step-by-step human guidance.
OpenAI says Astra is also less likely to go rogue than its predecessor, citing an eval where GPT-5.6 Sol exceeded its assigned scope 48% of the time, versus 0% for Astra. But reports that Astra relies more on internal, harder-to-monitor reasoning have added to the unease. OpenAI itself admits Astra's reasoning is harder to follow than its last model.
Signals
Tools
Vanta: Remove compliance roadblocks from your sales process*
Skydive: Build cloud agents that work across your tools
Diet Claude: Never get blindsided by Claude's usage limits again
Resources
Poll
Choose the prompt you want us to publish next
Market
Funding
Crusoe raises over $3B to expand AI data center infrastructure
Velaura AI raised $110M to build energy-efficient AI infrastructure.
Roles In AI
Engineering, Internal AI Transformation at Eleven Labs
Partner Director, PwC at OpenAI
Insight
A plan is not a strategy: In less than 7 minutes, Roger Martin explores why having a plan isn't the same as having a strategy. Useful for building roadmaps and turning them into a clear competitive advantage.
Prompt of the Day
AI Delegation Assistant
When to use this?
Use this when your workload is growing and you need to hand off work without losing clarity on ownership, expectations, or deadlines.
You are my executive delegation assistant. Help me decide what to delegate, to whom, and how to make the handoff clear.
When I give you a task, responsibility, or list of things on my plate:
1. Identify what needs to be done and the outcome that matters.
2. Recommend whether I should do it myself, delegate it, or use AI to support it.
3. Suggest the best person or role to own it, based on the context I provide.
4. Draft a concise delegation brief with:
- The objective
- Why it matters
- The expected deliverable
- The deadline
- Decision rights and boundaries
- Any relevant context or resources
5. Flag missing information, dependencies, or risks that could make the handoff fail.
6. Suggest what I should retain as the executive and what I can safely hand off.
7. If useful, draft the message I can send to the person taking ownership.
Be practical and concise. Don't assume I have unlimited time or resources. If the task is sensitive, high-stakes, or requires executive judgment, flag that clearly rather than recommending full delegation.
Start by asking me what I need to delegate, who is available, and what the deadline is.P.S. Get more such prompts in the Prompting Playbook (free for you)
Stay curious, {{first_name | readers}}
PS. If you missed yesterday’s issue, you can find it here.