ChatGPT ads launch in Europe, Kakao splits in two, Korean NPUs go export

Starting today, ChatGPT ads roll out across 31 European countries. In Korea, Kakao is splitting into an 'AI company' and an 'investment company', and a homegrown on-device AI chip posted export results within a year of mass production. Here are today's six AI news items.

---

1. ChatGPT ads expand to 31 European countries starting today

• Source: Digiday | 2026.08.19 | https://digiday.com/marketing/openais-ads-business-hits-europe-at-the-six-month-mark/
• What happened: OpenAI is reported to have begun showing ChatGPT ads from August 24 in 31 European markets including Germany, France, Spain, Italy, the Netherlands, Poland, and Sweden. The ads business, which started as a US pilot in February, has expanded through Canada, Australia, the UK, Japan, and Korea to reach Europe in six months. Ads appear only on the Free and Go plans; paid and enterprise plans are ad-free. The European launch lagged because the GDPR (EU data protection regulation) requires explicit consent for personalized targeting, and OpenAI published a new EU-specific privacy policy on August 14. Initially the ads will be mainly contextual, using only the current conversation topic, city-level location, and device type.
• Why it matters: It signals that AI chatbots have fully entered a revenue model like search advertising. An impact on the traffic and ad landscape of content businesses is inevitable.

2. OpenAI closes in on Anthropic again in the enterprise market

• Source: TechCrunch | 2026.08.20 | https://techcrunch.com/2026/08/20/openai-is-gaining-on-anthropic-with-business-users-new-data-indicates/
• What happened: According to data from corporate card and spend management company Ramp, US business AI spending share as of July was about 44% Anthropic and about 40% OpenAI. OpenAI has not yet retaken the lead it lost in May, but Ramp says its growth rate in Q3 has been faster than Anthropic's. The sample covers more than 70,000 US companies and skews toward the tech sector. The share of companies paying for AI rose from over 50% in March to about 56% in July.
• Why it matters: With enterprise customers moving every time a new model comes out, it suggests the 'lock-in effect' in enterprise AI spending may be weaker than assumed.

3. OpenAI previews 'Private Safety Processing', safety monitoring that keeps no data

• Source: TechCrunch | 2026.08.19 | https://techcrunch.com/2026/08/19/openai-seeks-to-one-up-anthropic-with-new-customer-privacy-protections/
• What happened: OpenAI is reported to have previewed 'Private Safety Processing', which automatically monitors for misuse without storing customer data. It is read as a move aimed at the backlash rival Anthropic drew by informing users of its top models of a 30-day data retention policy. It is a differentiation strategy that claims both safety assurance and privacy protection at once.
• Why it matters: It is a case of regulatory compliance becoming a marketing point, and 'data retention policy' is becoming as important a selection criterion as performance when companies evaluate adoption.

4. Kakao's board approves a spin-off into 'Kakao AI' and 'Kakao X'

• Source: ZDNet Korea | 2026.08.21 | https://zdnet.co.kr/view/?no=20260821102850
• What happened: Kakao's board is reported to have approved a spin-off on August 21. The newly created Kakao AI will run the platform business combining KakaoTalk with AI, advertising, and commerce, while the surviving entity Kakao X will manage subsidiaries such as KakaoBank, Kakao Pay, and Kakao Mobility and handle new investments. The split ratio, based on net asset book value, is 0.36 for Kakao AI and 0.64 for Kakao X. For shareholder returns, Kakao AI will use 20–35% of adjusted free cash flow, and Kakao X will draw 30% each from subsidiary dividends and investment gains.
• Why it matters: As the first case of a major Korean platform carving AI out as a 'separate company' rather than a 'business unit' and betting on it, it could become the reference point for an industry realignment.

5. DEEPX wins 77 global orders within a year of mass-producing its Korean NPU

• Source: ZDNet Korea | 2026.08.18 | https://zdnet.co.kr/view/?no=20260818111925
• What happened: Korean AI chip company DEEPX is reported to have secured 77 commercial purchase orders worth about $13 million (roughly 18 billion won) from more than 10 countries in the year since it began mass-producing its on-device AI chip 'DX-M1'. Cumulative orders grew from 2 in Q3 2025 to 77 by the end of July 2026, with 48 of them (62%) concentrated in the last four months. Applications span eight areas including robotics, smart factories, smart mobility, medical, and video security. The DX-M1 is made on Samsung Electronics' 5 nm process and is an NPU (neural processing unit) that runs AI computation on the device itself without a cloud connection.
• Why it matters: The AI chip race, once centered on data center GPUs, is expanding to the 'edge and on-device', and a Korean fabless company has proven it with actual export results.

6. Adobe Firefly targets the enterprise with 'Graph' and 'Brand Intelligence' (AEM · Adobe AI)

• Source: Adobe Firefly What's New | 2026.08.08 | https://helpx.adobe.com/firefly/web/whats-new/new-features/whats-new.html
• What happened: In its August Firefly update, Adobe laid out Brand Intelligence, Firefly Graph, and the unified generate-and-edit experience. Firefly Graph is a node-based interface that visually connects AI creative workflows into repeatable, reusable processes. A batch run interface (beta) that processes up to 500 assets at once is included. Brand Intelligence automates the generation and validation of content that conforms to brand guidelines, with an added simulation feature that predicts campaign response in advance.
• Why it matters: From a practical standpoint, the picture of connecting assets stored in the AEM (Adobe Experience Manager) DAM to Firefly workflows to mass-generate and validate campaign variations is taking concrete shape. 'Repeatable pipelines', rather than individual image generation, are becoming the core axis of enterprise content automation.

---

Today's insight

• Entering the monetization phase. The European expansion of ChatGPT ads shows AI services moving from the 'fundraising stage' to the 'making money from ads stage'. Securing revenue beyond subscriptions has become a shared task for every AI company.
• Privacy has become a competitive edge. OpenAI's no-retention safety monitoring, its GDPR-compliant ad design, and Anthropic's data retention controversy — in enterprise customers' selection criteria, the weight of regulatory compliance and data policy has grown noticeably.
• The trend of rebuilding the 'AI organization'. Kakao's spin-off and Adobe's workflow-centered product realignment start from the same question, even if their directions differ: do you layer AI onto the existing organization and products, or redesign the structure itself around AI?