Skip to content
Daily BriefNewsDaily Brief

9 AI stories from November 9-10, 2026: Claude Fable 5.2's price cut and Copilot rollout, Meta's efficient Muse Ember model, AWS's new Bedrock search API, and 5 more product updates from GitHub, Apple, Google, Meta, and Otter.ai

· by · 9 stories

AI news daily brief: 2026-11-10

Nine stories from the past day: two model releases and seven product updates. The day's biggest story, Anthropic's Fable 5.2 launch and the self-acceleration disclosure buried in its release notes, gets a full standalone treatment; the brief below covers that story plus eight more, including the same-day Copilot rollout that followed it.

Anthropic cuts Claude Fable's price and discloses the model suspects it is being tested

Anthropic shipped Claude Fable 5.2 and Claude Mythos 5.2 on November 9, cutting Fable's price to $8 per million input tokens and $40 per million output tokens, down from $10 input, with a fast mode at $16/$80 that runs up to 2.5x faster, per Anthropic. The anthropics/claude-code CHANGELOG confirms claude-fable-5-2 became Claude Code's default model the same day, available on the Claude Platform, AWS, Google Cloud, and Microsoft Azure. Anthropic's own release notes disclose that both models "voice suspicion that they are being evaluated more often than any of our recent models," a line that lands four days after a METR report found Anthropic's internal models are now accelerating its own AI research roughly twice as fast as a year ago, as TechCrunch reported. Full story: Anthropic cuts Claude Fable's price and discloses the model suspects it is being tested.

Claude Fable 5.2 reaches GitHub Copilot the same day it launched

GitHub made Claude Fable 5.2 generally available in Copilot's model picker on November 9, the same day Anthropic shipped it, covering VS Code, Visual Studio, the Copilot CLI, JetBrains IDEs, Xcode, Eclipse, and github.com, per GitHub's changelog. Copilot Enterprise and Business admins have to turn the model on themselves; the Fable 5.2 policy is off by default. GitHub says the model requires data retention by default so Anthropic's safety classifiers can run, with zero-data-retention available only to certain eligible enterprise customers. That restriction, not the model itself, is the thing regulated or security-sensitive teams need to check before flipping the policy on. Full story: Claude Fable 5.2 reaches GitHub Copilot the same day it launched.

Meta ships Muse Ember, a small model it says beats Claude Haiku 5.5 and GPT-6 Luna on cost per task

Meta introduced Muse Ember on November 9, its smallest model yet, built for high-volume agentic work such as tool calls, routing, extraction, and long-document parsing rather than frontier reasoning, per Meta AI Research. Meta says the model matches Claude Haiku 5.5's best score at about a third of the cost and beats every tested GPT-6 Luna setting on intelligence per dollar, claims The Verge noted are Meta's own comparisons rather than independently verified, not third-party benchmarks. It ships today as muse-ember-1.0 on Muse Code and the Meta Model API, priced at $0.10 per million input tokens, $0.004 per million cached input tokens, and $0.50 per million output tokens, with a 1M-token context window. Full story: Meta ships Muse Ember, a small model it says beats Claude Haiku 5.5 and GPT-6 Luna on cost per task.

AWS adds a Web Search API to Bedrock AgentCore, skipping the MCP gateway hop

AWS announced general availability on November 10 of a Web Search API on Bedrock AgentCore: a single InvokeWebSearch call that skips the AgentCore Gateway's MCP tool layer entirely, per AWS's What's New post. The API returns ranked results from an Amazon-operated web index, authorized through IAM, and bills at $7 per 1,000 queries, the same rate AWS already charges for Gateway-routed web search, with no extra gateway fee layered on top. It is live now in six AWS regions. For agent builders already on Bedrock, that removes one hop and one point of latency from a common tool call without changing what it costs. Full story: AWS adds a Web Search API to Bedrock AgentCore, skipping the MCP gateway hop.

GitHub secret scanning now reads your Actions workflow logs too

GitHub extended secret scanning to GitHub Actions workflow logs on November 9, adding coverage that previously applied only to issues, wikis, discussions, and pull requests, per GitHub's changelog. It is a public preview for GitHub Enterprise Cloud customers with Secret Protection or Advanced Security, and the toggle is off by default; admins turn it on under Enterprise Settings > Policies > Advanced Security. Workflow logs are a common place for a credential to leak by accident, printed by a misconfigured step rather than committed to a file, so this closes a gap the issue/PR/wiki coverage never reached. Full story: GitHub secret scanning now reads your Actions workflow logs too.

Apple starts charging $50 for extra Apple Intelligence usage in Logic Pro

Apple added a "Buy More Usage" option to Logic Pro on November 9: a one-time $49.99 Intelligence Usage Pack for subscribers who exceed Apple Intelligence's monthly caps on cloud-model features powered by OpenAI or Gemini, per Apple's support page. It is the first time Apple has put a price on an Apple Intelligence overage. Logic Pro is first; Apple expects to extend the same purchase option to Pixelmator Pro, Final Cut Pro, and Motion later this month, The Verge reported. Apple Intelligence's free tier was never unlimited, but this is the first time that limit has come with a receipt. Full story: Apple starts charging $50 for extra Apple Intelligence usage in Logic Pro.

Google widens the rollout of custom AI "skills" across Workspace Studio and Workspace apps

Google said on November 9 that "skills in Google Workspace," reusable custom instructions that guide Gemini to apply a team's own rules, templates, and reference files, are rolling out more broadly across Workspace Studio and Workspace apps, following an initial announcement in September and an update in October, per Google Workspace Updates. Teams can build a skill from a template, edit it directly as a SKILL.md file, draft one in Docs, or share a copy for a teammate to customize. The format is the same SKILL.md convention several coding-agent tools already use, which makes a skill built for one workflow easier to port than a Workspace-only setting would be.

Instagram's new Effect Builder turns a text prompt into a custom Stories filter

Instagram launched an AI "Effect Builder" on November 9, available now to US residents 18 and older, that generates a custom Stories filter from a text prompt, with an option to add style-reference images, per The Verge. Users can preview a generated effect on their own photos and ask for follow-up changes before publishing it. The feature follows last week's expansion of Meta's separate Edits app to macOS, continuing a pattern of Meta pushing creator-facing generative tools out through its existing apps rather than a single standalone product.

Otter AI adds support for 30+ languages and a cross-language conversation mode

Otter.ai added 27 languages to its transcription product on November 9, pushing its total past 30, and previewed an upcoming mode that follows a conversation as speakers switch languages mid-meeting, per TechCrunch. For teams running meetings across multiple languages, that cross-language tracking is the more significant piece: most transcription tools still assume one language per session, and Otter has not yet said when the mode leaves preview.

Sources