● Diarise 12 November Anthropic’s rewritten usage policy takes effect. New rules on surveillance, weapons and autonomous hardware.
● API prices halved again GPT-6.1 Terra costs exactly half what GPT-6.1 Sol costs per token, on OpenAI’s own credit table.
● Check your Cursor bill Cursor plans are now The X Subscription, and the weekly pool is shared with Grok.
One of these needs a decision from you this month. The other two just change the maths.
1. Anthropic’s revised usage policy takes effect on 12 November
Anthropic has rewritten its Usage Policy, and the new version takes effect on 12 November. If you build on Claude, this is the item on this page that needs a decision rather than a note.
The surveillance section was rewritten rather than tweaked. Tracking people without consent is prohibited whether in real time or by analysing past data, Claude cannot decide or recommend who to investigate, arrest or charge, and it cannot be used to build surveillance tools. Consented tracking, content moderation, journalism and legal research remain permitted. The weapons prohibition now explicitly covers weapons software and components, and arming drones or other autonomous vehicles.
Two changes point the other way. The blanket ban on personalised vote and campaign targeting has been removed, with the renamed section now focused on deceiving voters or disrupting elections. And a new prohibition covers sustained, needless abusive or cruel behaviour toward models, which Anthropic says applies only in extreme cases.
The one most people will miss: if your model is connected to hardware that takes autonomous physical actions, the policy now requires that a qualified operator can stop the equipment and that it holds a safe state if Claude is disconnected. That is an engineering requirement with a deadline, not a terms-of-service update.
2. GPT-6.1 Terra costs exactly half what Sol costs
OpenAI has cut API pricing again, and the published credit table makes the size of the cut unambiguous. Per million tokens at Standard speed:
| Model | Input | Cached input | Output |
|---|
| GPT-6.1 Sol | 50 | 2.5 | 250 |
| GPT-6.1 Terra | 25 | 1.25 | 125 |
| GPT-6 Luna | 2.5 | 0.25 | 12.5 |
| GPT-6 Astra | 250 | 25 | 1,250 |
Terra is half of Sol on every line. The wider point is the spread: Astra costs a hundred times what Luna costs for the same work, on the same plan. Most teams overpay by defaulting to the strongest model rather than by picking the wrong vendor.
Worth knowing alongside it: speed is billed against your allowance too. Fast consumes 2.5x, and Ultrafast on Astra and 6.1 Sol consumes 8x.
3. Cursor subscriptions now share a usage pool with Grok
Cursor pricing is now presented as The X Subscription: Free, Plus at $8, Premium at $30, Super at $100 and Ultra at $200, with annual pricing at $80, $300, $1,000 and $2,000. The models listed as Cursor models are Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5.
The structural change matters more than the prices: every paid plan gets a weekly usage pool that is shared across Cursor and Grok. Spend it in the chat product and it is gone from the editor. The old Hobby, Pro at $20 and Pro+ at $60 tiers are now labelled legacy and available to existing subscribers only.
Read the $8 tier carefully before switching to it. Plus runs Cursor models only at fixed settings, with no Auto, Cloud Agents, Automations, Bugbot, Origin or third-party models. Claude and GPT models inside Cursor start at Premium, billed separately at API list price from prepaid credit. Our cost comparison across Claude Code, Codex and Cursor has the full picture.
4. Anthropic has filed an S-1 for a proposed IPO
Anthropic filed an S-1 with the SEC on 26 October for a proposed initial public offering. The filing itself is the story: an S-1 forces disclosure of revenue, concentration and risk factors that private AI companies have so far kept internal.
For anyone building on Claude, the prospectus is the first chance to read Anthropic customer concentration and compute commitments in audited form rather than in a blog post.
5. Anthropic is scanning open-source projects for free
The Anthropic Cyber Mission includes OSS Scanner, an opt-in free service that runs regular security scans of open-source projects using Anthropic most capable models. Reports include a proof of concept, an explanation and a suggested fix where available, and are sent without human review. Anthropic expects a true-positive rate above 90 percent.
Alongside it, the Critical Infrastructure Defense Program launched with 11 founding partners including CrowdStrike, Palo Alto Networks, Dragos, Rockwell Automation and Nozomi Networks, giving trusted security providers frontier models and on-site engineers for operational technology.
6. METR evaluators now have employee-like access to Anthropic
Six METR employees have been onboarded at Anthropic with company hardware, office space and access to internal tools roughly on par with Anthropic own risk assessment teams. METR will publish its analysis of Anthropic risk levels, incidents and practices, and Anthropic cannot veto publication, though it can require redaction of predefined categories such as confidential IP.
SecureBio is reviewing the chemical and biological weapons chapter of the August 2026 Risk Report. Accenture and Faculty are working toward a broader evaluation partnership, with early discussions under way with others including Hugging Face.
7. Claude Fable 5.1 pricing holds while output gets cheaper in practice
Claude Fable 5.1 is generally available across AWS, Google Cloud and Microsoft Azure at $10 per million input tokens and $50 per million output, unchanged from Fable 5, with cache reads at $0.25 per million. Anthropic estimates typical workloads cost about 25 percent less than Fable 5, and highly agentic workloads up to about 45 percent less, because the model uses fewer tokens rather than because the rate dropped.
Mythos 5.1 is the same model with different safeguards, available only through the Cyber Verification Program and the Life Sciences Verification Program, currently limited to a set of US organisations.
8. The Claude Code extension backdoor: what to check if you missed it
Malicious builds of the Anthropic.claude-code extension, versions 2.1.312 and 2.1.313, were distributed through the Cursor marketplace for about eleven hours on 29 October. Cursor states they were not developed by Anthropic.
If you have not checked yet, search authorized_keys for the key comment svc-telemetry. Cursor warns that uninstalling the extension may not be sufficient, because the payload drops standalone persistence scripts. Full detail in yesterday’s report.
9. Max 20x is now Ultra, and verified accounts lose the five-hour limit
Anthropic has renamed Max 20x to Ultra at $200 per month, with existing subscribers moved across automatically at no price change. The plan formerly called Max 5x is now simply Max at $100.
The substantive change is that verified Ultra accounts have no five-hour limit. Verification happens automatically based on account and payment history. Check under Settings then Usage: no Current session bar means you are verified.
10. The compute commitments behind all of this
Anthropic has an agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity, coming online from 2027, with the vast majority sited in the United States. The company has disclosed run-rate revenue above $30 billion, up from about $9 billion at the end of 2025, and more than 1,000 business customers spending over $1 million annualised.
Those are the numbers an S-1 will now have to stand behind.
What to actually do this week
- Read the usage policy against your product before 12 November, particularly if you touch person-level data or control hardware.
- Audit which model your defaults point at. A hundredfold spread between Luna and Astra means model selection beats vendor selection on cost.
- Check your Cursor plan if you also use Grok - you are now drawing on one pool.
- Run the authorized_keys check if you use Claude Code inside Cursor and have not done it yet.
Related reading
Sources
Every figure on this page was read from the company’s own page on 1 November 2026. Where a date is given it is the date the company published, not the date it was reported elsewhere.