2026.11.05 AI news briefing — the memory supercycle and the agent war
Contents
Anthropic unveiled a technology that cryptographically proves an AI model's 'identity,' and at the same time refiled IPO paperwork showing trillion-won-scale revenue. In Korea, Naver, Kakao, and Baemin are moving beyond search to accelerate competition in action-taking AI agents, and Samsung Electronics posted record-high results on the back of the memory supercycle.
1. Anthropic unveils technology to prove an AI model's 'identity'
- Source: TechCrunch | 2026.11.03, Help Net Security | 2026.11.04
- Key details: Anthropic's Frontier Red Team unveiled an open-source prototype called 'FlashProve.' The technology cryptographically verifies whether a specific model actually generated a specific response, publishing a per-model 'fingerprint' that lets responses be checked against it without exposing the model's weights. It was tested on a 400-billion-parameter mixture-of-experts (MoE) model, running at about 110 tokens per second on an H100 GPU, and the proving and verification code was released on GitHub under an Apache 2.0 license.
- Why it matters: This could become a verification tool against the practice of secretly swapping in a cheaper model behind an API, or using a premium model only for benchmarks. That said, it's still early-stage, with no concrete plans yet for production use.
2. Anthropic files amended S-1 ahead of IPO — $17 billion in Q3 revenue
- Source: CNBC | 2026.11.02
- Key details: In an amended S-1 filed with the SEC, Anthropic disclosed third-quarter revenue of about $17 billion, up roughly 1,127% year over year, with nine-month cumulative revenue of $33.2 billion. Nine-month net loss, however, reached $104 billion, driven by fair-value changes in convertible notes and derivatives. Total compute and infrastructure purchase commitments and leases amount to $895.4 billion.
- Why it matters: Growth remains steep, but the quarter-over-quarter growth rate (47%) slowed from Q2 (144%). OpenAI is also reportedly preparing a confidential IPO this year, suggesting the trillion-dollar-scale fundraising race between the two leading AI companies is heating up.
3. Naver, Kakao, and Baemin accelerate the shift from 'search' to 'action' with AI agents
- Source: Electronic Times | 2026.11.04
- Key details: Naver has moved up the timeline for its 'integrated agent,' originally targeted for next year, to aim for the first half of 2027. The goal is to combine its search, shopping, and maps agents into one that handles everything seamlessly from search to booking and payment. Kakao plans to start a private beta of 'Kanana Space,' an integrated AI assistant built on KakaoTalk, in December, and Baedal Minjok (Baemin) has been running a private beta since the 20th of last month for an AI search feature that recommends restaurants using natural language.
- Why it matters: This signals that the axis of competition among Korean platforms is shifting from 'AI that finds information well' to 'AI that performs tasks on your behalf.'
4. Samsung Electronics posts 107 trillion won in Q3 operating profit — first global tech firm to top 100 trillion won
- Source: Yonhap News | 2026.10.29
- Key details: Samsung Electronics posted record results for the fourth consecutive quarter, with Q3 revenue of 195.3 trillion won and operating profit of 107.4 trillion won. Its Device Solutions (DS) semiconductor division drove most of the results with 106.5 trillion won in operating profit, on the back of surging demand for AI data center memory and the ramp-up of HBM4 mass production. Of its annual capital expenditure of 84.4 trillion won, 76.8 trillion won went to the semiconductor division.
- Why it matters: The figures show the AI infrastructure investment boom feeding directly into the earnings of Korean semiconductor companies.
5. Alibaba, Tencent, and ByteDance seek '10-year' long-term contracts from Samsung and SK Hynix
- Source: Crypto Briefing | 2026.10.27
- Key details: China's three largest cloud companies have reportedly asked Samsung Electronics and SK Hynix for long-term agreements (LTAs) to supply server DDR5 and low-power LPDDR5 memory through 2035. That's an unusually long term, given that LTAs typically run three to five years. The request comes after US Big Tech firms already locked up 70–80% of the two companies' 2027 production, and the three Chinese firms' recent Nvidia GPU orders alone exceed $35 billion combined.
- Why it matters: Memory supply shortages have emerged as a new bottleneck in the global AI infrastructure race, and Korean memory makers' bargaining power is growing accordingly.
Today's takeaway
- The center of gravity in AI competition appears to be shifting from 'smarter models' toward 'trustworthy verification' (FlashProve) and 'securable infrastructure' (memory and capital).
- Korean platforms are moving the battlefield away from search competition toward agents that actually perform tasks.
- The memory supercycle has become a variable that simultaneously shapes Korean semiconductor companies' earnings and global AI companies' procurement strategies.
References
Contents
Related posts

2026.10.09 AI news briefing — OpenAI revenue dispute and Samsung’s quarterly operating profit of 100조
Amid a sell-off in AI stocks over disputes about OpenAI’s actual annual revenue, Samsung Electronics surpassed a quarterly operating profit of 100조원 for the first time on booming memory demand.
2026.10.19 AI news briefing — AI safety warnings and the spread of AI services in Korea
OpenAI disclosed more rogue-agent incidents; Anthropic strengthened Claude's election safeguards. Toss Bank expanded AI consultation to all customers; Kakao set a 5M MAU target for 'AI for All.'
2026.10.17 AI news briefing — OpenAI state lawsuits and Samsung-LG's AI chip clash
Five top AI stories: U.S. states suing OpenAI over child safety, scientists' open letter for AI oversight, Anthropic's safety transparency push, and Samsung and LG's AI chip and industrial exhibits.

2026.09.23 AI News Briefing — Claude Opus 5.5 and GPT-6 Price Cuts, and AI Safety Coordination
Anthropic and OpenAI cut prices on new models in tandem, and Google, Anthropic, and OpenAI unveiled cybersecurity safeguards on the same day. The UN Security Council put AI risk on its agenda for the first time, and in Korea, Rebellions' major funding round and the delivery-app AI race round out today's stories.