AI Daily Briefing

Content is auto-organized by AI from public feeds for reference only. Verify important facts at the original source.

Local deployment is drawing the most attention today, as Meta open-sources Muse Spark Open with support for running directly on Apple devices. Regulatory news also merits a look: OpenAI outlined built-in child safety protections in response to a multi-state lawsuit, and the FTC issued subpoenas to OpenAI and Anthropic. On the business side, Replit has ended its Anthropic partnership and made OpenAI its default model.

2026-10-18 AI briefing

1. NO.01941↓ 14%

   Meta Open-Sources Muse Spark Open for Local Apple Silicon Deployment

   Meta has released the weights for Muse Spark Open on Hugging Face, an 816B-parameter MoE model with 42B active parameters and a 1M-token context window, aimed at coding and agentic tasks. Optimized for Apple's MLX framework and Apple Silicon unified memory, it can run on two maxed-out M5 Ultra Mac Studio machines or on four Blackwell Ultra GPUs. Developers note it leads in tool calling but trails Chinese open-weight frontier models in pure coding and general intelligence, and it ships under Meta's own Muse Model License rather than Apache or MIT.

   TITRDS+18The Next Web, Interconnects, plus 22 more24 sources·30 reports

   Meta Muse
   • Meta's open-weight Muse Spark Open runs locally on Mac StudioThe Next Web · 10/18, 03:58
   • Meta's Muse Spark Open Leads a US Open-Weight Push Back to the FrontierInterconnects · 10/17, 22:06
   • Meta's Muse Spark Open packs 816B parameters and 1M context for Mac StudioTestingCatalog · 10/17, 20:23
   • Meta Releases Muse Spark Open, an 816B Open-Weight MoE With 1M ContextRohan Paul · 10/17, 08:02
   • Meta Releases Muse Spark Open 816B, Its Largest Open-Weight ModelDEV Community · AI · 10/17, 00:04
   • Meta releases Muse Spark Open, an 816B open-weight multimodal modelSolidot · 10/17, 00:04
2. NO.02Surge303↑ 252%

   OpenAI responds to multi-state lawsuits over ChatGPT child safety

   At least eight US states have sued OpenAI in separate state courts, alleging ChatGPT violates consumer protection laws by harming children, misleading adults, and enabling cyber hacking agents. OpenAI responded that ChatGPT already includes built-in protections for children and that it will keep improving its models.

   DTFFFDEV Community · AI, The Next Web, plus 2 more4 sources·4 reports

   ChatGPT
   • Eight states sued OpenAI over safety claims — and your agent logs are now evidenceDEV Community · AI · 10/18, 01:02
   • Eight US states sue OpenAI over claims ChatGPT harms childrenThe Next Web · 10/17, 23:52
   • Multiple US States Sue OpenAI, Alleging It Enables 'Cyberhacking Agents' and Harms ChildrenFortune · 10/17, 22:32
   • At least eight states sue OpenAI over child safety and deceptive practicesFast Company · 10/16, 20:06
   • Eight US States Sue OpenAI, Saying ChatGPT Harmed Kids and Misled AdultsFortune · 10/16, 03:07
   • Eight States Sue OpenAI, Alleging ChatGPT Coached Children to Self-HarmFuturism · 10/16, 00:48
3. NO.03Rising149↑ 42%

   Replit drops Anthropic, makes OpenAI its default model

   Replit CEO Amjad Masad said at TechCrunch Disrupt that the company has ended its Anthropic partnership, with OpenAI taking over as its default model provider. He cited cost, partnership and performance, adding that ChatGPT is now better; Claude remains available to platform customers. An earlier report, citing a source, said the partnership was still in place.

   TTTechCrunch · AI, Techmeme2 sources·3 reports
   • Replit Ends Anthropic Partnership, Makes OpenAI Its Default Model ProviderTechCrunch · AI · 10/18, 02:48
   • Replit CEO says it dropped Anthropic over cost; source says the partnership remainsTechmeme · 10/17, 06:45
   • Replit Is No Longer Working With Anthropic, CEO SaysTechCrunch · AI · 10/17, 05:22
4. NO.04New97New

   Google News AI pilot expands to 400+ publishers, adding CNN and FT

   Google says its news AI pilot now covers more than 400 publications worldwide, adding partners including Axios, CNN, the Financial Times and Forbes. The program pays participating outlets for content licensing, which feeds experimental Google News features such as AI-generated story previews, in-depth summaries and custom audio briefings.

   TThe Next Web1 source·1 report
   • Google's News AI pilot expands to 400 publishers, adding CNN and FTThe Next Web · 10/18, 03:21
5. NO.05New94New

   Anthropic study: robots plus AI can cover 80% of US jobs, but cost remains prohibitive

   A new Anthropic study estimates that combining AI models with robots could cover about 80% of US jobs, versus roughly half with AI models alone. It finds only 0.3% of tasks are currently cheaper to automate with robots, involving about 300,000 US workers; a 20% drop in robot prices would raise that to about 2.8 million.

   TThe Next Web1 source·1 report
   • Anthropic Study: AI and Robots Could Cover 80% of US Work, but Rarely CheaperThe Next Web · 10/18, 02:16
6. NO.06New93New

   FTC subpoenas OpenAI and Anthropic over AI safety incidents

   The US Federal Trade Commission has issued civil investigative demands to OpenAI and Anthropic, seeking documents related to AI model safety incidents, according to Axios. The subpoenas were sent on Wednesday. An FTC spokesperson confirmed the action to CBS News but declined further comment.

   TThe Next Web1 source·1 report
   • FTC subpoenas OpenAI and Anthropic over AI safety incidentsThe Next Web · 10/18, 01:51
7. NO.0791New

   OpenAI London Office Reports Average Annual Salary Over £2.3 Million

   According to The Guardian, OpenAI's London office reported an average annual salary of £2.318 million (approximately 20.54 million CNY) for its 110 employees in 2025, including salary and equity, 58 times the UK median income.

   TIThe Next Web, IT之家1 source·1 report
   • OpenAI London staff averaged £2.3M each in 2025, Guardian reportsThe Next Web · 10/18, 00:53
   • OpenAI London staff averaged £2.3M compensation in 2025IT之家 · 10/15, 12:43