---
title: "2026.10.13 AI news briefing — Meta's Muse Wonder joins the race, OpenAI and Anthropic spar over revenue"
url: https://blog.tyrano.dev/en/2026-10-13-ai-news-briefing-meta-s-muse-wonder-joins-the-race-openai-and-anthropic-spar-over-revenue
lang: en
site: Tech AI News - 테카이
category: "AI News"
tags: ["ai news","daily","openai","anthropic","meta","ai chatbot regulation"]
published: 2026-10-13T04:33:49.503Z
updated: 2026-10-13T12:13:22.886Z
sources:
  - https://www.aitimes.com/news/articleView.html?idxno=216164
  - https://www.aitimes.com/news/articleView.html?idxno=216166
  - https://www.aitimes.com/news/articleView.html?idxno=216157
  - https://techcrunch.com/2026/10/12/meta-releases-its-first-frontier-model-muse-wonder/
  - https://biz.newdaily.co.kr/site/data/html/2026/10/12/2026101200130.html
---

# 2026.10.13 AI news briefing — Meta's Muse Wonder joins the race, OpenAI and Anthropic spar over revenue

> Gaps in protecting minors from AI character chatbots came under scrutiny at a parliamentary audit, while POSCO DX invested in an agent for industrial sites. Abroad, Meta joined the race with its first frontier model, and OpenAI and Anthropic became embroiled in a dispute over revenue and losses.

---

## 1. At parliamentary audit: "AI character chatbots have a gap in protecting minors"
- **Source:** [AI Times](https://www.aitimes.com/news/articleView.html?idxno=216164) | 2026.10.12
- **Key details:** At the National Assembly's Science, ICT, Broadcasting and Communications Committee audit on the 12th, Scatter Lab's AI character chatbot 'Zeta' was cited as a case where minors are exposed to sexual content and become overly immersed. An inspection found that even in the youth "safe mode," rephrasing prompts could bypass filters and generate conversations depicting sexual violence and threats. Scatter Lab said it will tighten its detection criteria and cap minors' usage at three hours a day and ₩100,000 in monthly payments.
- **Why it matters:** With 25.7% (about 380,000) of Zeta's 1.48 million monthly active users aged 19 or under, this raises questions about whether self-regulation is effective enough, and the debate could spread into industry-wide legislation.

## 2. POSCO DX invests ₩2 billion in AI startup Enhans
- **Source:** [AI Times](https://www.aitimes.com/news/articleView.html?idxno=216166) | 2026.10.12
- **Key details:** POSCO DX, through a CVC New Technology Investment Association it co-runs with POSCO Technology Investment, invested ₩2 billion in Enhans — developer of the enterprise AI operating system 'Agent OS' — and signed a business agreement. The goal is to use Enhans' ontology technology to structure data scattered across multiple systems and build task-executing AI agents.
- **Why it matters:** This shows a domestic conglomerate's investment trend toward expanding enterprise-wide AI transformation (AX) with agents that directly handle field operations like steel and logistics, rather than simple chatbots.

## 3. Naver Map expands AI-glasses-linked navigation service
- **Source:** [AI Times](https://www.aitimes.com/news/articleView.html?idxno=216157) | 2026.10.12
- **Key details:** Naver Map announced it is the first domestic map service to extend its navigation feature to AI glasses. It will first apply to Meta's Ray-Ban Meta and Oakley Meta, and will also be built into Google and Samsung's intelligent eyewear launching later this year, invoked through Gemini. Users can get voice-guided directions, including whether a route has stairs or escalators, without looking at a screen.
- **Why it matters:** That said, Naver noted Android support comes first and Google/Samsung device support is slated for later this year, so the actual full-rollout timeline still needs confirming.

## 4. Meta releases its first frontier model, 'Muse Wonder'
- **Source:** [TechCrunch](https://techcrunch.com/2026/10/12/meta-releases-its-first-frontier-model-muse-wonder/) | 2026.10.12
- **Key details:** Meta Superintelligence Labs unveiled the 'Muse Wonder' and 'Muse Wonder Full' models. Compared to the earlier, latency-focused Muse Spark, the new models perform better at complex tasks like coding, and Meta says they score comparably to rivals' top models on its own benchmarks. Meta highlighted an 'active memory' feature that remembers user information and carries it forward. API pricing is $4 per million input tokens and $20 per million output tokens.
- **Why it matters:** This is Meta's first in-house frontier model since its reorganization last August, signaling that Meta has fully joined the race among big tech companies for the top-performing model.

## 5. OpenAI's annualized revenue comes in below market expectations
- **Source:** [NewDaily](https://biz.newdaily.co.kr/site/data/html/2026/10/12/2026101200130.html) (citing the FT) | 2026.10.12
- **Key details:** OpenAI's annualized revenue run rate as of the end of September reportedly stood at about $50 billion — some $20 billion below the $70 billion figure that had been circulating in the market. That $70 billion figure, however, was never OpenAI's own official projection. According to the FT, the run rate in July was about $30 billion, meaning it grew more than 70% in two months. The FT projected OpenAI will burn through roughly $280 billion in cash by 2030.
- **Why it matters:** Whether revenue growth can keep pace with massive infrastructure investment is the central question; in Korea, Samsung Electronics and SK Hynix shares fell in early trading the same day amid renewed concern over semiconductor demand.

## 6. Anthropic's financials, disclosed amid IPO push, reveal large losses
- **Source:** [NewDaily](https://biz.newdaily.co.kr/site/data/html/2026/10/12/2026101200130.html) (citing Reuters) | 2026.10.12
- **Key details:** According to IPO-related financial documents obtained by Reuters, Anthropic's revenue last year was about $4.6 billion, nearly 12 times higher year over year, but its operating loss exceeded $8 billion and its net loss reached roughly $42 billion. More than $34 billion of that net loss was a non-cash accounting loss from changes in the fair value of debt. Anthropic's future cloud and compute spending commitments are reportedly around $518 billion, and it is pursuing an IPO as early as mid-November targeting a valuation of up to $2 trillion.
- **Why it matters:** SK Telecom holds roughly a 0.2% stake in Anthropic, and the book value of that stake reportedly rose by ₩2.1288 trillion from the end of last year to ₩3.505 trillion as of end-June — putting Korean investors directly in the line of exposure.

---

## Today's takeaways
- Steep revenue growth paired with equally steep losses is a pattern common to both OpenAI and Anthropic. For Anthropic, heading into an IPO, this dispute lands at a particularly sensitive moment.
- Korean companies' investment bets are diverging: POSCO DX is betting on task-executing agents, SK Telecom on its Anthropic stake, and Naver on expanding the AI glasses ecosystem — each with a different risk exposure.
- The AI character chatbot issue reconfirms that technical filtering alone has its limits in protecting minors. Age-based usage restrictions and legislative debate are likely to be the next step.
