AI Board

AI BOARD

AI Board

AI tools, policy moves, and public decisions are tracked in plain language.

All posts AI Briefs Tools Policy Education Industry

AI Daily Briefing 2026-06-15

Author
김 경진
Date
2026-06-15 06:14
Views
160
AI Daily Briefing 2026-06-15
2026-06-15 · KIMKJ.COM AI DAILY BRIEFING

Global AI Daily Briefing

Monday, June 15, 2026

Top 3

1. The U.S. Department of Commerce imposed export controls on Anthropic's Fable 5 and Mythos 5. Global access was cut off just 72 hours after launch. The trigger was Amazon CEO Andy Jassy reporting security vulnerabilities to the White House, and White House adviser David Sacks said, "Anthropic refused to make the fixes."

2. The G7 Leaders' Summit opens today in Evian, France. This is the first time the heads of three AI companies, Altman (OpenAI), Amodei (Anthropic), and Hassabis (DeepMind), will attend at the same time.

3. Salesforce Summer '26 is officially released today. Agentforce multi-agent orchestration moves out of beta and becomes generally available.

AI Technology Shifts

U.S. Commerce Department Orders Export Controls on Anthropic Fable 5 and Mythos 5

Fable 5, unveiled on June 9, disappeared within 72 hours. That is unprecedented. Commerce Secretary Howard Lutnick sent an export-control letter at 5:21 p.m. ET on June 12, and Anthropic, unable to filter foreign users in real time, cut off access for all customers worldwide.

The chain of events began on June 10. Jailbreak researcher "Pliny the Liberator" disclosed on X that Fable 5's safety classifier had been bypassed through Unicode tricks and multi-agent decomposition. The claim was that cyberattack information, explosives, and chemical synthesis pathways had been extracted. Amazon's security research team independently confirmed the same vulnerability, and CEO Andy Jassy called Treasury Secretary Scott Bessent directly.

Anthropic says this is a "misunderstanding." It argues that the same level of jailbreak is possible with other public models, including OpenAI GPT-5.5. Claude models other than Fable 5 and Mythos 5 are unaffected. The restoration date is unknown.

(CNBC, 2026-06-12 / TechCrunch, 2026-06-12 / TechTimes, 2026-06-14)

Reports of an "Alignment Tax" After the Fable 5 Block

After Fable 5 was taken down, users began reporting lower accuracy in existing Claude models. The phenomenon in which performance drops when the level of safety alignment is raised has been labeled an "alignment tax." There is no paper-level verification yet.

(TechTimes, 2026-06-14)

AI Law / Legislation

G7 Evian Summit Opens June 15, First-Ever Simultaneous Attendance by Three AI CEOs

French President Emmanuel Macron sent Altman a direct invitation. The agenda includes the voluntary AI code of conduct, the environmental impact of data centers, and the United States' AI stack export policy. The summit runs through the 17th. Because it comes right after the Fable 5 export controls, AI security is expected to carry extra weight in the discussions.

(Bloomberg, 2026-06-12 / EU Council)

[Follow-up] EU AI Act Omnibus Agreement Delays High-Risk System Enforcement to December 2027

Under a provisional agreement reached on May 7, mandatory enforcement for high-risk systems, including hiring, credit scoring, and law-enforcement AI, has been pushed from August 2026 to December 2027. AI embedded in regulated products now faces a new deadline of August 2028. Fines for prohibited practices, including social scoring, up to 35 million euros or 7% of revenue, are already being enforced. This corrects earlier reports that said implementation was imminent on August 2.

(Travers Smith / Inside Privacy, 2026-06)

Administrative Oversight / Regulatory Investigations

Amazon CEO Jassy Reveals How the Fable 5 Export Controls Were Triggered

Amazon CEO Andy Jassy called Treasury Secretary Scott Bessent and senior officials on June 12 and relayed the Fable 5 vulnerability findings uncovered by Amazon's security research team. The export-control letter was issued right after that call. Amazon is Anthropic's largest investor, with about $8 billion across the full series, and its cloud host through AWS Bedrock. Because it is acting as investor, infrastructure provider, and the party that triggered regulation, conflict-of-interest questions are rising.

(TechTimes, 2026-06-14)

David Sacks: "Anthropic Refused to Fix the Vulnerability"

David Sacks, co-chair of the President's Council of Advisors on Science and Technology, disclosed the sequence on June 14. He said the administration told Amodei to "either fix the vulnerability or take the model down," but was refused. Sacks added, "As soon as the fix is complete, we want to lift the export controls as quickly as possible." According to Semafor, concerns over Chinese access to Mythos were also part of the backdrop.

(Benzinga, 2026-06-14 / Tom's Hardware, 2026-06-14 / Semafor, 2026-06-13)

New Technology / Product Launches and Practical Use Cases

Salesforce Summer '26 Official Release, Agentforce Multi-Agent Orchestration Goes GA

It becomes officially available starting June 15. Multi-agent orchestration, where multiple AI agents operate like a team to handle complex workflows, has moved from beta to GA. The IT service domain pack comes with more than 50 specialized AI agents built in, automatically detecting and handling employee request intent.

(Salesforce, 2026-06-15)

U.S. AI Trend Check

Technology: Export controls on Fable 5 and Mythos 5 cut off global access just 72 hours after launch. The Department of Commerce moved after a jailbreak researcher's public disclosure coincided with a report from Amazon's security team. After the block, accuracy declines have also been reported in existing Claude models.

Law / Regulation: Sacks said Anthropic's refusal to make fixes was the reason for the export controls. The G7 Evian Summit opens today, with Altman, Amodei, and Hassabis attending. Because it comes right after the export controls, AI security is expected to be a central agenda item.

Products / Use Cases: Salesforce Summer '26 moves to GA today, and Agentforce multi-agent orchestration becomes an official service.

China AI Trend Check

Law / Regulation: According to Semafor's June 13 report, one of the reasons behind the export controls was that Chinese users had been accessing Anthropic's top-tier model Mythos without restriction. The White House viewed that as a national security concern. Anthropic's lack of a technical mechanism to distinguish U.S. use from overseas use was decisive.

Attorney Kim Kyung-jin's Commentary

The Anthropic Fable 5 episode is the first case in which the federal government has imposed export controls on a commercially deployed AI model and blocked access worldwide.

Amazon sits at the center of the matter. It is the largest investor and the provider of cloud infrastructure through AWS Bedrock, and this time it also reported the security vulnerability to the White House and triggered the export controls. A structure in which one company serves at once as investor, host, and the party that triggers regulation will have a hard time avoiding conflict-of-interest controversy. From Anthropic's position, its largest investor effectively reported the risks of its own model to the government.

Anthropic's pending IPO cannot be left out either. If export controls drag on, they will directly affect valuation. Sacks' statement expressing hope for a swift lifting of the controls can be read as a sign that the administration understands that burden as well.

A precedent has now been set. Once the government has used its power to take down a commercial AI model, the same logic can be applied to other models. How quickly a company responds to jailbreak vulnerability reports has become a standard that can divide legal risk.

[Reference Links]

Official statement on suspension of access to Anthropic Fable 5 and Mythos 5

https://www.anthropic.com/news/fable-mythos-access

Amazon CEO triggered Fable 5 export controls (TechTimes, 2026-06-14)

https://www.techtimes.com/articles/318350/20260614/amazon-triggered-claude-fable-5-shutdown-investor-cloud-host-now-regulator.htm

David Sacks on lifting the Fable 5 controls (Benzinga, 2026-06-14)

https://www.benzinga.com/markets/tech/26/06/53185788/sacks-us-wants-to-lift-anthropic-export-controls-as-soon-as-possible-after-fable-5-fix

How Anthropic refused to fix the vulnerability (Tom's Hardware, 2026-06-14)

https://www.tomshardware.com/tech-industry/artificial-intelligence/trump-adviser-david-sacks-says-anthropic-refused-to-fix-fable-5-jailbreak-before-us-export-controls

Report on Claude alignment tax (TechTimes, 2026-06-14)

https://www.techtimes.com/articles/318357/20260614/claude-accuracy-degradation-after-fable-ban-has-name-alignment-tax.htm

Background concerns over Chinese access to Mythos (Semafor, 2026-06-13)

https://www.semafor.com/article/06/13/2026/white-house-move-to-limit-anthropic-linked-to-concerns-about-chinese-access-to-mythos

AI CEO attendance at the G7 Evian summit (Dataconomy, 2026-06-12)

https://dataconomy.com/2026/06/12/ai-leaders-openai-google-deepmind-anthropic-g7-summit/

Official G7 Evian agenda (EU Council)

https://www.consilium.europa.eu/en/meetings/international-summit/2026/06/15-17/

Delay in EU AI Act Omnibus enforcement (Travers Smith)

https://www.traverssmith.com/knowledge/knowledge-container/eu-agrees-to-delay-key-ai-act-compliance-deadlines/

Salesforce Summer '26 release (Salesforce, 2026-06-15)

https://www.salesforce.com/news/stories/summer-2026-product-release-announcement/

Global shutdown under Fable 5 export controls (CNBC, 2026-06-12)

https://www.cnbc.com/2026/06/12/anthropic-disables-access-to-fable-5-and-mythos-5-to-comply-with-government-directive.html

How the Fable 5 jailbreak unfolded (TechCrunch, 2026-06-12)

https://techcrunch.com/2026/06/12/anthropics-safety-warnings-may-have-just-backfired-the-government-has-pulled-the-plug-on-its-most-powerful-ai/

Search keywords: AI news June 15 2026, OpenAI Anthropic Google Microsoft AI June 15 2026, 中国 人工智能 June 2026, AI regulation law lawsuit June 15 2026, G7 summit AI CEOs Altman Amodei Hassabis June 15 2026, Anthropic Fable 5 Mythos 5 export control, Amazon triggered Fable 5 shutdown, David Sacks Fable 5, alignment tax Claude, Salesforce Summer 26 Agentforce, EU AI Act Omnibus high risk delay, ChatGPT Claude Gemini new feature June 2026, 百度 阿里 腾讯 字节 AI 2026年6月, Korea Japan AI news June 2026

Generated at: 2026-06-15 06:00 KST

What matters is not how hard it is, but whether it solves the problem. In the end, that is planning. The real moat is not code, but experience, trust, and data.

📚 Source Materials — click to expand
KIMKJ.COM
#김경진 #김경진변호사 #김경진인공지능 #인공지능 #AI #AI전문가 #AI법률 #AI정책 #AI규제 #AI윤리 #생성형AI #ChatGPT #Claude #GPT #LLM #디지털전환 #스마트시티 #자율주행 #데이터규제 #GDPR #개인정보보호 #AI거버넌스 #국회의원김경진 #법률전문가 #테크정책 #AI교육 #AI행정혁명 #AI패권전쟁 #kimkj #kimkjcom
kimkj.com Home
Scroll to Top
kimkj.com Home kimkj.com Home