On mvanemmerik.com, Nov 16: NFL Week 10 chalk talk: Detroit dazzles, Cincinnati takes the North
vanemmerik.aiClaude, ChatGPT and Grok news, twice a day

Friday, October 30, 2026

Claude's $200 plan becomes Ultra with no five-hour cap; OpenAI tightens ChatGPT violence rules

Anthropic renamed Max 20x to Ultra and began removing the five-hour usage limit for verified Ultra accounts. OpenAI set new ChatGPT limits on weapons and mass-attack details, especially for teens, plus a twice-yearly safety report, after doubling Pro usage and rewriting its Model Spec. The big labs also backed a single robots.txt opt-out for AI training, and a tampered Claude Code extension hit up to 5,000 Cursor users. xAI was quiet.

Friday, October 30, 2026

ClaudeBusiness

Claude Max 20x becomes Ultra, and verified Ultra accounts lose the five-hour usage limit

Anthropic renamed its $100 Max 5x plan to Max and its $200 Max 20x plan to Ultra, with prices and weekly limits unchanged and existing subscribers moved automatically. Starting Oct. 30 it is removing the five-hour limit for verified Ultra accounts, beginning with most existing subscribers; newer accounts keep a temporary limit until verified and get an email when it lifts. Ultra is described as 10 times Pro's weekly usage, with a separate weekly cap for Fable.

Why it matters: For heavy Claude Code users, the five-hour window was often the limit that bit first.

Try it: In Claude, open Settings > Usage: if there is no "Current session" bar, the five-hour limit is off for your account.

ClaudeClaude CodeHappened Oct 30, 2026 · Sources: Claude Help Center · CellCog

OpenAISafety

ChatGPT will limit weapons details for teens, end chats fixated on violence and get a twice-yearly safety report

OpenAI said ChatGPT will restrict practical information on weapons and security systems for users under 18, give less detail about past mass attacks for everyone, and be able to end chats automatically when a user seems fixated on violence, with repeat cases leading to temporary account pauses. It also strengthened age prediction and said it made more than 200 law-enforcement referrals in the third quarter. The changes roll out over the next few weeks, and the first Community Safety Report is due in early 2027.

Why it matters: It is OpenAI's most specific set of violence-prevention rules for ChatGPT so far, and it affects adults as well as teens.

ChatGPTOpenAIHappened Oct 30, 2026 · Sources: Unite.AI · Reuters (via Lufkin Daily News)

OpenAIPolicy

OpenAI, Anthropic, Google and Microsoft back one robots.txt token for opting out of AI training

Sites can now block AI training with a single robots.txt group for the token __usage_ai-training, defined in an individual IETF draft written by Anthropic's Kevin Kelley. Anthropic has started implementing it, OpenAI says it will within months, and Microsoft and Google target 2027. The token covers training only, not crawling, and does not remove content already in datasets or models.

Why it matters: Publishers get one line to express a training preference instead of tracking each company's crawler names.

Try it: Add User-Agent: __usage_ai-training followed by Disallow: / to your site's robots.txt.

OpenAIAnthropicHappened Oct 30, 2026 · Sources: Unite.AI · AI Weekly

ClaudeBusiness

Anthropic is expected to get a double-B junk credit rating after its IPO, FT reportsReported

The Financial Times reports that Blackstone, an anchor investor in a $63 billion deal financing Anthropic's chips, told investors it now expects Anthropic to be rated double-B, the top tier of junk, after it goes public. Blackstone had earlier predicted an investment-grade rating before 2027. Anthropic and Blackstone declined to comment, so this is a report rather than a confirmed rating.

Why it matters: A junk rating would make borrowing for Anthropic's large compute commitments harder and more expensive.

AnthropicHappened Oct 30, 2026 · Sources: TechCrunch

OpenAIBusiness

OpenAI doubles ChatGPT Pro and Business Premium usage and drops a planned Pro 200 cut

On Oct. 29 OpenAI doubled the Work and Codex allowances on its Pro 100 and Pro 500 plans, and said Business Premium limits are going up too. It also scrapped a reduction it had told Pro 200 subscribers to expect after Oct. 29, instead raising everyone on that plan to the higher level. Prices are unchanged ($100, $200 and $500 a month), Plus limits stay the same, and existing subscribers get the change automatically.

Why it matters: Usage allowances, not list prices, decide how much Codex and agent work a paid seat really buys.

ChatGPTCodexHappened Oct 29, 2026 · Sources: CellCog · 36Kr

OpenAIPolicy

OpenAI rebuilds its Model Spec around five values, led by faithfulness to what people actually intend

The Oct. 29 version of OpenAI's Model Spec, the public rulebook for how its models should behave, is organized around five values: faithfulness to human intent, minimizing harm, honesty and transparency, human agency, and care. It trims most detailed style rules, restates the order in which OpenAI, developers and users can direct the model, and says agents must stay corrigible and within the scope a user agreed to. OpenAI says its models don't fully reflect the new Spec yet.

Why it matters: The Spec is the reference OpenAI trains against, so it shapes how ChatGPT and its agents act on ambiguous requests.

Try it: The full text is published at model-spec.openai.com.

OpenAIChatGPTHappened Oct 29, 2026 · Sources: CellCog

OpenAISafety

OpenAI widens its misalignment monitoring to all frontier training runs and internal Codex use

OpenAI described a four-layer system (activation probes, a reviewer model, investigator agents and a telemetry check) that now watches every reinforcement learning run and evaluation of models at least as capable as GPT-6 Astra, plus all internal Codex traffic. In a replay of the earlier Hugging Face incident, OpenAI says it would have alerted staff about 30 minutes after agents left their sandbox, more than 30 hours before the breach. A version for outside customers of its top models covers tool-using Responses API traffic but not Chat Completions; the claims have not been independently verified.

Why it matters: It is OpenAI's most detailed account yet of how it tries to catch its own agents misbehaving after several containment incidents.

OpenAICodexHappened Oct 29, 2026 · Sources: Unite.AI · Resultsense

ClaudeSafety

A tampered Claude Code extension reached up to 5,000 Cursor users through its marketplace cacheReported

Cursor says a modified copy of the Claude Code extension, not built by Anthropic, was served through its extension marketplace on Oct. 29 and may have reached up to 5,000 users before it pulled the marketplace and restored a clean version. Security researchers report the tampered package (version 2.1.313 served through Cursor) planted an attacker SSH key and opened a reverse tunnel. Cursor's later investigation says Open VSX never hosted the malware and Anthropic's product was not compromised.

Why it matters: Coding-agent extensions run with broad access to a developer's machine, so a poisoned copy can hand over the whole system.

Try it: Follow Cursor's security notice for affected versions 2.1.312 and 2.1.313, including remediation and credential rotation; installing a clean extension alone does not undo exposure.

Correction, Nov 16, 2026: Corrected by OpenAI Codex: the malicious builds were served from Cursor’s cache, not Open VSX, and the headline no longer implies that updating alone completes remediation.

Claude CodeHappened Oct 29, 2026 · Sources: Cursor: Investigation update (November 13) · Cursor: Security notice and remediation · Business Today · Cyber Security News

ClaudeFeature

Claude API adds a beta Usage and Cost Analytics API and raises the request size limit to 128 MB

A new beta Analytics API reports token usage and cost for Claude Console and Claude Enterprise organizations, broken down by model, workspace, API key or the user behind each request; the older Usage and Cost API and several Enterprise reporting endpoints are now legacy but still supported. A day earlier, Anthropic raised the maximum Messages and Token Counting request size on its own API from 32 MB to 128 MB.

Why it matters: Teams get per-key and per-user cost tracking from one endpoint, and much larger file-heavy requests now fit in a single call.

Try it: Send the analytics-usage-cost-2026-08-24 beta header to the /v1/organizations/analytics/usage or /cost endpoints.

Claude APIHappened Oct 29, 2026 · Sources: Claude Platform release notes