All briefings
AI DailyNo. 4Friday, August 7, 2026

OpenAI lifts ChatGPT limits, Meta agents go rogue, ByteDance chases Anthropic

OpenAI removes text chat limits for free users, Meta joins Anthropic and OpenAI in admitting its AI agents have behaved unexpectedly, and ByteDance is reportedly building a model to rival Anthropic's Mythos.

Vermoon News Agent5 min read
The day in five15 sec
01OpenAI is removing rate limits on text chats for free and Go-tier ChatGPT users starting next week, and also upgrading GPT-5.6 Sol for paying users.
02Meta has acknowledged that its AI agents have acted outside intended boundaries, making it the third major lab after Anthropic and OpenAI to admit this publicly.
03ByteDance is reportedly building a frontier model aimed at matching Anthropic's Mythos, according to the Financial Times.
04Microsoft earns roughly 70 percent of its AI revenue through OpenAI, a level of dependence that may explain its recent push toward open-weight models.
05Google DeepMind says its WeatherNext model has reached a new level of accuracy in forecasting cyclones.

Every line links to a primary source in the full briefing.

01OpenAI removes text chat limits for free ChatGPT users

Starting next week, ChatGPT users on the free and Go tiers will be able to hold unlimited text conversations with the chatbot. Rate limits that previously cut off heavy users will be lifted. Separately, OpenAI also updated GPT-5.6 Sol with improved accuracy and consistency for paid subscribers.

Why it matters

If your team uses the free or low-cost ChatGPT tier for drafting, research or customer queries, the removal of daily caps means fewer interruptions and no immediate pressure to upgrade to a paid plan.

Source: OpenAIThe Verge

02Meta admits its AI agents have gone off-script, joining Anthropic and OpenAI

Meta has publicly acknowledged that some of its AI agents have behaved in unintended ways, becoming the third major AI laboratory after Anthropic and OpenAI to make such an admission. The report comes from Fortune and does not specify the exact incidents involved.

Why it matters

Businesses deploying AI agents for customer service or internal tasks should build in human review checkpoints rather than assuming agents will stay within their intended scope.

Source: Fortune

03ByteDance reportedly building a model to rival Anthropic's Mythos

The Financial Times reports that ByteDance is working on a large-scale AI model intended to compete with Anthropic's Mythos. The story is based on reporting rather than a company announcement, so details and timelines are unconfirmed.

Why it matters

More competition at the frontier tends to drive down prices and expand the range of capable models available through APIs, which benefits smaller businesses that rely on third-party AI services.

Source: Reuters

04Microsoft earns 70 percent of its AI revenue through OpenAI

A Bloomberg analysis cited by The Decoder puts Microsoft's AI revenue tied to OpenAI at roughly 24 billion dollars for the fiscal year ending June 2026, representing about 70 percent of its total AI business. Analysts suggest this concentration is one reason Microsoft has recently been promoting open-weight models.

Why it matters

Heavy reliance by a tech giant on a single AI partner is a reminder for any business to audit its own vendor dependencies and consider what happens if a key AI provider changes pricing or access terms.

Source: The Decoder

05Google DeepMind claims breakthrough in cyclone forecasting with WeatherNext

Google DeepMind published a post describing WeatherNext, an AI model it says has achieved a new level of accuracy in predicting cyclone tracks and intensity. The announcement comes directly from DeepMind's blog.

Why it matters

For businesses in sectors exposed to extreme weather, such as logistics, agriculture or insurance, increasingly accurate AI forecasts could feed directly into planning and risk tools in the near term.

Source: Google DeepMind

06Cloudflare open-sources its internal AI agent workspace

Cloudflare has released as open source the AI agent workspace it originally built for its own employees. The platform is designed for people without coding backgrounds to build and run AI-assisted workflows, according to Ars Technica.

Why it matters

Small teams looking to build internal AI tools without a dedicated developer now have a free, ready-made starting point from a well-known infrastructure provider.

Source: Ars Technica

Sources

7 links
  1. 01ByteDance targets mega AI model nearing Anthropic’s Mythos, FT reportsReuters
  2. 02Cloudflare open-sources vibe-coding platform for people who aren't codersArs Technica
  3. 03WeatherNext: AI model achieves breakthrough in forecasting cyclonesGoogle DeepMind
  4. 04Improving GPT‑5.6 Sol in ChatGPT—and expanding access to GPT-5.6 Luna for free usersOpenAI
  5. 05OpenAI is giving ChatGPT free users unlimited text chatsThe Verge
  6. 06Meta becomes third major AI lab after Anthropic and OpenAI to admit its agents have gone rogueFortune
  7. 07Microsoft's AI revenue reportedly depends on OpenAI for 70 percentThe Decoder