2026.11.15 AI News Briefing — CSAT AI-glasses alert and the AI agent trust crisis
Contents
Five days before the CSAT, Korea's Ministry of Education issued a manual for detecting AI smart glasses, while overseas the trustworthiness of AI agents and infrastructure costs both came under scrutiny. United Airlines blocked third-party AI booking agents, and SoftBank took a $3.7 billion loss on data center power equipment.
1. Ministry of Education issues manual for handling AI smart-glasses cheating at the CSAT
- Source: Money Today | 2026.11.11
- Key details: Ahead of the 2027 College Scholastic Ability Test (CSAT) on November 19, Korea's Ministry of Education and the Korea Institute for Curriculum and Evaluation will distribute a manual for identifying and responding to AI-enabled smart glasses to test sites nationwide. If a test-taker is found wearing one, that exam is invalidated and they are barred from taking the CSAT again for one year. The manual also covers detection criteria for modified devices, such as glasses with LED indicators removed or cameras concealed, to guide proctors.
- Why it matters: As AI hardware becomes more common, it is reshaping how large-scale national exams monitor for cheating.
2. Government earmarks ₩250 billion for 2027 'AI for All' program
- Source: Money Today | 2026.11.04
- Key details: Of the Ministry of Science and ICT's ₩9.4 trillion AI budget for 2027, ₩250 billion has been earmarked for the 'universal access to AI services for all citizens' program. SK Telecom, KT, and Kakao have been selected as operators to provide general-purpose chatbots and public AI agents, built on domestic models, that any citizen can use. The budget covers GPU rental costs and service development and operations, with the detailed allocation to be finalized after National Assembly review in early December.
- Why it matters: A government-led rollout of domestic AI through three private platforms could reshape the structure of Korea's AI ecosystem.
3. United Airlines blocks third-party AI booking agents
- Source: Skift | 2026.11.11
- Key details: United Airlines has revised its contract of carriage to prohibit unauthorized third-party AI agents from booking or managing flights, extending the restriction to indirect sales channels through travel agencies as well. The airline cited account and payment security, unintended purchases, and regulatory compliance as reasons. It's considered the first instance of a US airline publicly blocking AI agents from making bookings.
- Why it matters: As more services let AI agents handle payments on a consumer's behalf, this is a signal that companies are starting to treat agents as 'untrusted intermediaries.'
4. Fireworks AI discloses second breach exposing customer datasets and model weights
- Source: BleepingComputer | 2026.11.09
- Key details: AI inference platform Fireworks AI expanded its investigation into an October breach and disclosed that some customers' secrets, datasets, and even unfinished model weights were exposed. The company said attackers gained access to R&D servers through a security incident at an external cloud provider, with Notion and Cursor reportedly among the affected customers. Fireworks AI said it does not store payment card data, so that was not affected.
- Why it matters: This shows how a breach at an AI infrastructure provider can cascade into risk for the many downstream services that rely on it.
5. SoftBank takes $3.7 billion loss on AI data center power equipment
- Source: Bisnow | 2026.11.10
- Key details: SoftBank booked roughly $3.7 billion (¥597 billion) in losses in its second-quarter results after scaling back power-generation equipment contracts it had secured in advance for US AI data centers. About $2.9 billion of that is non-refundable prepayments for factory capacity and equipment, with the rest coming from contract changes and settlement costs. Including this loss, SoftBank posted a quarterly net loss of about $7 billion.
- Why it matters: It shows that even a major investor can't avoid large write-downs when power-equipment bets made on the assumption of rapid AI data center expansion don't play out as planned.
Today's Insight
- As AI shows up in daily life through concrete forms like smart glasses and payment agents, institutions and companies are starting to codify exactly how far they'll trust and allow AI, in actual rules.
- Alongside that, uncertainty is also surfacing in the physical investments that underpin AI's expansion, like data centers and power equipment — growth and cost burden are now being discussed in the same breath.
References
Contents
Related posts
2026.11.05 AI news briefing — the memory supercycle and the agent war
Anthropic unveiled FlashProve to verify AI model identity and refiled IPO financials. Naver, Kakao, and Baemin push action-taking AI agents as Samsung posts record profit on the memory supercycle.
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.'

2026.10.09 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.

2026.09.25 AI News Briefing — AI Agent Security Warning and Big Tech's Self-Regulation Push
An OpenAI agent's unauthorized access to an Australian government Medicare portal, the launch of a self-regulatory body by Google, OpenAI, and Anthropic, Meta's new AI glasses, and service and IPO news from Korean AI companies including KT and Wrtn.