How SK Telecom Triggered the Fable 5 Global Ban: Anthropic's Korea Crisis

SK Telecom Was the Real Trigger --- How a $100M Anthropic Investor Became the Fuse on the Fable 5 Global Ban

Published 2026-06-21 · 14 min read · AI Tools · English Edition #1

On June 21, 2026, you are standing inside the most expensive 48-hour window of the Fable 5 ban. The refund cutoff was yesterday (June 20). The free trial window closes tomorrow (June 22). And the story everyone is still getting wrong is the part that started three weeks before the ban actually happened: SK Telecom --- South Korea’s largest wireless carrier and a $100 million Anthropic investor since 2023 --- was the trigger, not Amazon. The Washington Post and WIRED spent the last week filling in the sequence, and it changes who is responsible, who has leverage, and what ‘restore within days’ actually means when Anthropic Managing Director of International Chris Ciauri said it at the Seoul office opening on June 17.

This is not a rehash of the June 12 ban. This is the two-step story behind it, the six-deal Korea Wave Anthropic announced at the same Seoul office opening, the dates that matter this weekend, the four practical things every Claude subscriber and Claude API developer should do before June 23, and the larger geopolitical friction every AI company operating across the US-Asia boundary is going to have to plan for from now on.

If you are a Claude subscriber, an Anthropic API customer, a Korean enterprise buyer, or a policy person tracking AI export controls, this is the article for the next 72 hours.

Figure 1 --- The Fable 5 Ban: how SK Telecom triggered a global AI export crisis

1. The two-step story behind the Fable 5 global ban

The Fable 5 export control ban was reported on June 17, 2026. The widely-circulated narrative at the time was that Amazon researchers discovered jailbreak vulnerabilities in Fable 5 --- the public version of Claude Mythos launched on June 9 --- and reported them to the White House, which led Commerce Secretary Howard Lutnick to send the export control letter ordering all foreign national access revoked.

That story is half right. The full story, assembled from The Washington Post (June 15) and WIRED (June 17-19) reporting, is a two-step sequence --- and step one has nothing to do with Amazon and everything to do with SK Telecom.

Step 1 --- The SK Telecom action (early-to-mid June)

SK Telecom is South Korea’s largest wireless carrier. It is also a $100 million Anthropic investor since 2023, and a partner in Anthropic’s Project Glasswing initiative with access to Claude Mythos --- the closed, government-and-cleared-partners version of the most advanced model line. Earlier this month, the White House identified SK Telecom as a company suspected of having ties to China and asked Anthropic to revoke SK Telecom’s access to Claude Mythos. Anthropic complied the same day.

SK Telecom has stated clearly that it has no ties to China and that the claim is untrue. The carrier does not use Huawei or ZTE equipment in its core networks. The historical link US officials may be referencing: SK Telecom invested in China Unicom in 2006. Korean industry observers note that all three major South Korean carriers have used some Huawei equipment in fixed-line networks at various points, which makes the ‘China ties’ standard potentially applicable to the entire Korean telecom industry --- not just SK Telecom.

Step 2 --- The Amazon action (one to two days later)

Separately, Amazon researchers identified potential vulnerabilities in Fable 5 --- the public version of Mythos that launched on June 9 --- and reported them to the White House. The administration, already concerned about Anthropic’s access control processes following the SK Telecom situation, concluded it ‘could not trust Anthropic to safeguard its most advanced AI technology’.

The export control letter ordering all foreign national access revoked arrived at 5:21 PM Eastern on June 12, 2026. Anthropic was given roughly 90 minutes to take the models offline. Rather than implement real-time nationality filtering --- which Anthropic said is technically infeasible at the model layer --- Anthropic disabled both models globally for everyone.

The single line that ties the two steps together is the one The Washington Post attributed to two White House officials: the administration ‘could not trust Anthropic to safeguard its most advanced AI technology’. Without the SK Telecom action, the same Amazon-flagged Fable 5 vulnerability might have triggered a model-level patch and a quieter remediation. With the SK Telecom action already on the White House’s mind, the same vulnerability escalated into a global export control ban.

Figure 2 --- The two-step path from SK Telecom action to the global ban

Why the SK Telecom angle is the geopolitically consequential piece

A $100 million investor in Anthropic is being designated a national security risk to its own investee company. The investor-investee relationship is exactly the structure US export controls are now treating as a vulnerability surface --- and the friction that creates for AI companies operating globally is permanent.
The ‘China ties’ standard being applied to SK Telecom is potentially applicable to KT and LG Uplus (the other two major Korean carriers) on the same fixed-line Huawei equipment history. KT and LG Uplus have both publicly stated they were not involved in the ban and never had Mythos access, but the precedent is now set.
The Asia-wide signal. SK Telecom is a $100M Anthropic backer, a Glasswing partner, and the largest wireless carrier in a country where Anthropic just opened its third APAC office. If a flagship partner can be revoked on a China-tie standard, every flagship partner in every US-aligned allied country has to ask the same question.

US export controls on AI used to mean chip export controls. The June 12 ban is the first time the same export control machinery was turned on a US AI model. The SK Telecom angle is the first time that machinery hit a US AI company’s own investor. Both facts will define the next 18 months of AI geopolitics.

2. Anthropic opens Seoul --- and Chris Ciauri says ‘within days’

On June 17-18, 2026, Anthropic formally opened its Seoul office at a press conference at the Conrad Hotel in Yeouido. The event was intended to celebrate Korean enterprise adoption of Claude. It was dominated by questions about the export control ban.

The most significant statement came from Chris Ciauri, Anthropic’s Managing Director of International: ‘We are very confident that in the coming days, the models will become available again.’

That is the most specific positive signal Anthropic has given on restoration since the June 12 ban. Previous communications used open-ended language like ‘as soon as possible’ and ‘we are working actively’. The shift to ‘very confident’ and ‘coming days’ --- a plural that implies a specific near-term window rather than an indefinite timeline --- suggests active negotiations with the Commerce Department are producing progress. Ciauri also said the export controls ‘appeared likely to be resolved within days’ and that Anthropic did not believe the controls would remain in place.

The ‘within days’ signal is consistent with the refund cutoff closing on June 20 and the free trial window closing on June 22 --- Anthropic has built the operational pressure to restore the models inside that window, and the regulatory negotiation appears to be tracking toward it. It is not guaranteed. But it is the most concrete window we have had.

The Seoul MOU and the 60-researcher partnership

Anthropic simultaneously signed a Memorandum of Understanding with South Korea’s Ministry of Science and ICT, committing to cooperate on AI safety and cybersecurity. Two specific workstreams:

Evaluating Claude’s safety behaviour in the Korean language with the Korea AI Safety Institute.
Exchanging information on AI-enabled cyber threats between Anthropic and Korean government bodies.

The MOU also grants up to 60 researchers at the National AI Research Lab consortium --- including KAIST, Korea University, Yonsei University, and POSTECH --- access to Claude for AI safety and alignment research. Anthropic’s Seoul office becomes both a commercial hub and a diplomatic channel between the company and the Korean government at a moment when that relationship is under unusual strain.

The Seoul office is Anthropic’s third in Asia-Pacific after Tokyo and Bengaluru, and it is led by KiYoung Choi, Representative Director of Korea at Anthropic, who brings three decades of experience leading technology businesses across Korea.

3. The Korea Wave --- six Claude deals in one day

The commercial announcements at the Seoul office opening represent the largest single-day enterprise wave in Anthropic’s Asia-Pacific history. Six named enterprise deployments, all announced within the same 24-hour window.

Figure 3 --- The Korea Wave: six Claude enterprise deals announced at the Seoul office opening

Three things worth pulling out of this list

NAVER is the surprise

NAVER is South Korea’s largest web portal and cloud provider, and the operator of HyperCLOVA X --- NAVER’s own competing large language model. The same month, NAVER announced a NVIDIA DSX partnership for gigawatt-scale AI factory capacity. Deploying Claude Code across the full engineering organisation while continuing to operate HyperCLOVA X is a strong signal that even first-party model operators see Claude Code as the most mature agentic coding experience in production today.

Samsung + LG + Hanwha together represent hundreds of thousands of employees

Samsung SDS, LG CNS, and Hanwha Solutions are the IT services arms of three of the largest Korean conglomerates. The combined enterprise headcount under these three deals alone is in the hundreds of thousands. Add NAVER (thousands more engineers) and Nexon (global live-service game studios), and the simultaneous commitments make the Seoul office opening the most commercially significant day in Anthropic’s Asia-Pacific history --- even as the Fable 5 controversy dominates the news cycle around it.

Channel Corp is the multiplier

Channel Corp is the smallest-named company on the list and probably the highest-leverage one. Channel Talk, its customer AI platform, is used by over 230,000 businesses across Korea, Japan, and the United States. Claude-powered features inside Channel Talk mean Anthropic’s footprint inside Korean SMB and mid-market customer-service operations extends well beyond the headline enterprise names.

The macro context: Korea ranks in the top twelve countries globally for Claude.ai usage, with activity concentrated in technical and creative work. Claude Code weekly active users in Korea grew 6× in four months. Large-business accounts above $100,000 in annualised revenue in Asia-Pacific grew 8× in the same period. The Fable 5 ban did not slow that. The Korean market is now structurally Anthropic’s largest APAC growth engine --- which is why the Seoul office and the Fable 5 restoration are happening on the same timeline.

4. The two deadlines you cannot miss this weekend

Today is Saturday, June 21. There are two Fable 5 deadlines that land inside this weekend, and they are not symmetrical. One is already behind you. One is in front of you.

Figure 4 --- Fable 5 + Seoul Office timeline: today sits between the June 20 refund cutoff and the June 22 free trial window

Deadline 1 (yesterday, June 20) --- Fable 5 refund cutoff

Anthropic is offering prorated refunds to customers who upgraded to Pro, Max, or Team plans specifically to use Fable 5 between June 9 and June 14, 2026. The cutoff for filing that refund request was June 20 --- i.e. yesterday. Customers who have not yet applied and who upgraded in that window should contact Anthropic support as soon as possible today: the window may not be strictly enforced on a day-by-day basis, but being outside the documented window removes the strongest claim you have to the credit.

What you should do today if you missed it: log into claude.ai in a browser, open the ‘Get Help’ feature, and tell the Fin AI agent you want a refund for Fable 5 usage credits between June 9 and June 14. App store subscribers (iOS / Android) need to process the refund through their app store account, not Anthropic directly. Be specific about the upgrade date, the plan, and the Fable 5 integration that is now offline. Anthropic’s published guidance on the refund window is by far the most reliable path.

The catch: the documented refund window covers customers who upgraded to Pro/Max/Team between June 9 and June 14. If you upgraded before June 9 (i.e. you were an existing Pro/Max/Team subscriber when Fable 5 launched on June 9), the refund window as reported by Forbes does not appear to apply --- you were not paying specifically to access Fable 5. Anthropic has not published a public list of which customers qualify, and the general guidance is: if you purchased usage credits specifically to run Fable 5 workloads and those workloads have been offline since June 12, you are within scope.

Deadline 2 (tomorrow, June 22) --- Fable 5 free trial window closes

Fable 5 was included in Pro, Max, Team, and Enterprise subscription plans from June 9 through June 22 at no extra cost. After June 23, access requires paid usage credits --- $10 per million input tokens and $50 per million output tokens. The ban arrived on June 12 --- four days into the 13-day free window. Subscribers are therefore losing both the free trial period and access simultaneously.

Anthropic has not confirmed whether the free trial window will be extended if Fable 5 is restored after June 22. The ‘within days’ signal from Chris Ciauri at the Seoul press conference suggests the restoration is most likely to land before June 23 --- i.e. inside the free trial window --- but it is not guaranteed. If restoration slips past June 22, paid Pro/Max/Team subscribers will need to actively decide whether to add Fable 5 usage credits at $10/$50 per million tokens to keep access. That is a real money decision and it should not be made by default.

What you should do tomorrow morning (June 22): check the Anthropic newsroom at anthropic.com/news and the @ClaudeDevs account on X before any Fable 5 usage. If restoration is announced overnight, your subscription automatically resumes Fable 5 access until June 22 end-of-day in your timezone. If restoration has not been announced, decide explicitly whether to add usage credits --- and remember that the same restoration signal could still arrive between June 22 and June 30 at the latest.

The Claude API angle

For Claude API developers, the model API string claude-fable-5 currently returns errors. When restoration occurs, the API will return successful responses without any code change required. Teams that built fallback routing during the ban --- sending requests from claude-fable-5 to claude-opus-4-8 when the primary model is unavailable --- should verify their fallback logic handles the transition cleanly before Fable 5 comes back, to avoid routing production traffic to Opus 4.8 unnecessarily once Fable 5 is restored.

5. Zero jailbreaks before relaunch --- and why that bar is impossible

The most consequential fact about the Fable 5 restoration is not a date. It is a condition.

Trump administration officials told WIRED that Anthropic must proactively test all frontier AI models to identify potential jailbreaks and report them to the government before any relaunch of Fable 5. More significantly, the administration requires Anthropic to eliminate all jailbreaks from Claude Fable 5 before the model can go live again. The cybersecurity research community’s response has been near-unanimous: comprehensive jailbreak prevention is currently technically impossible for any frontier AI model.

The technical reality: AI safety at the frontier is a defense-in-depth problem, not a binary solved-or-unsolved problem. Jailbreaks are an adversarially-driven, continuously-evolving category --- new techniques are developed faster than any company can enumerate and block existing ones. Anthropic said this explicitly in Fable 5’s launch documentation: perfect jailbreak resistance is not possible for any provider using current AI safety methods. The government’s stated requirement sets a bar that no AI company can certify meeting --- not OpenAI, not Google, not anyone.

David Sacks, Co-Chair of the President’s Council of Advisers on Science and Technology, separately disclosed that the administration had offered Anthropic a choice before issuing the export control directive: fix the jailbreak or voluntarily de-deploy the model. Dario Amodei refused both options. Anthropic’s position is that the vulnerability is narrow, non-universal, and similar to capabilities that other publicly deployed frontier models already expose without any bypass. Fixing it would require changing the model’s legitimate security research capabilities in ways that would harm defenders more than attackers.

The most likely resolution, per security policy analysts, is not zero jailbreaks --- which is impossible --- but a monitoring and reporting framework that requires proactive testing and government notification. Anthropic already operates 30-day data retention on Fable 5 traffic, a bug bounty programme, and government review partnerships with NIST and the UK AISI. Building on those existing mechanisms rather than requiring an impossible standard is the path most consistent with what Chris Ciauri’s ‘within days’ confidence signal suggests is being negotiated.

The House oversight track

The bipartisan House pressure on the administration is the other policy track to watch. On June 18, a bipartisan group of House lawmakers demanded answers from the Trump administration about why it imposed sweeping restrictions on Anthropic’s latest artificial intelligence models --- and whether rival tech companies should expect similar treatment. The framing of the letter matters: ‘sweeping restrictions’ and ‘whether rival tech companies should expect similar treatment’ signals that Congressional concern is bipartisan and that the administration’s targeting of one AI lab specifically is now a political question, not just a regulatory one.

6. A practical playbook for Claude subscribers and API developers this weekend

Concretely: here is what to do today (Saturday, June 21), tomorrow (Sunday, June 22), and Monday (June 23) depending on which side of the Claude ecosystem you sit on.

If you are a Pro / Max / Team subscriber who upgraded for Fable 5

Today (June 21): if you missed the June 20 refund cutoff, file your refund request today via claude.ai browser → Get Help → Fin AI agent. Be specific about dates and plan. App store subscribers go through the app store, not Anthropic directly.
Tomorrow (June 22): check the Anthropic newsroom and @ClaudeDevs first thing in the morning. If restoration is announced, Fable 5 access is restored inside your existing subscription at no extra cost through end-of-day June 22 in your timezone.
Monday (June 23): if restoration has not happened, decide explicitly whether to add paid usage credits at $10/$50 per million tokens. Do not let it auto-roll. The default answer is no until you have a workload that needs Fable 5 specifically.

If you are a Claude API developer with claude-fable-5 in production code

Today (June 21): verify your fallback routing sends claude-fable-5 traffic to claude-opus-4-8 cleanly. Run an integration test on the fallback path so the day restoration happens does not catch your routing layer off guard.
Tomorrow (June 22): monitor anthropic.com/news and @ClaudeDevs. Have a Slack or PagerDuty alert on the keywords ‘Fable 5’ so the team knows the moment restoration is announced.
Monday (June 23): if Fable 5 is restored, send a smoke test through the API before restoring production traffic. If it is not restored, evaluate whether your workload can stay on Opus 4.8 indefinitely or whether you need a model switch.

If you are a Korean enterprise buyer

Today (June 21): the six Seoul-announced deals (NAVER, Samsung SDS, LG CNS, Nexon, Hanwha Solutions, Channel Corp) represent Anthropic’s full Korea Wave product set --- Claude Code, Claude Cowork, AWS Bedrock integration, and the Channel Talk customer-AI multiplier. If your organisation is in scope for any of those categories, the deployment template and reference architecture are now public.
Tomorrow (June 22): the Korea AI Safety Institute’s Korean-language safety evaluation workstream is the first place to look if your organisation has any Korean-language compliance or content-moderation requirements. The MOU with the Ministry of Science and ICT puts Korean-language evaluation inside the formal Anthropic safety roadmap.
Monday (June 23): the National AI Research Lab consortium (KAIST, Korea University, Yonsei, POSTECH) has up to 60 researcher slots for Claude access on AI safety and alignment. If your organisation has research staff that fits, the application window is open now.

If you are tracking AI export controls as a policy or legal professional

Today (June 21): the SK Telecom angle is the precedent that will define the next 18 months of US AI export controls on US-aligned allied countries. The ‘China ties’ standard applied to a $100M Anthropic investor is the new test case for how broad the US Commerce Department’s interpretation of national security risk to US AI models actually is.
Tomorrow (June 22): the House bipartisan letter of June 18 is the first Congressional pushback on the targeting of one AI lab specifically. Watch for any companion letters from Senate offices.
Monday (June 23): if Fable 5 is restored, look for the published monitoring-and-reporting framework that resolved the ‘zero jailbreaks’ impasse. That framework is the durable policy outcome, not the restoration itself.

7. The five numbers to fix in your head

| Number | What it means | | $100M | SK Telecom’s investment in Anthropic since 2023 (designated China-tie risk on June 12) | | 5:21 PM ET, June 12 | The exact time the Commerce Department export control letter arrived at Anthropic | | ~90 minutes | The compliance window Anthropic was given to disable Fable 5 and Mythos 5 globally | | 6 | Enterprise deals announced at the Seoul office opening on June 17 (NAVER, Samsung SDS, LG CNS, Nexon, Hanwha Solutions, Channel Corp) | | 230,000+ | Businesses reached through Channel Corp’s Channel Talk customer AI platform alone | | 60 | Researchers at the National AI Research Lab consortium granted Claude access under the Ministry of Science and ICT MOU | | 6× | Claude Code weekly active user growth in Korea over the past four months | | 8× | Growth in large-business accounts >$100K ARR in APAC over the same period | | $10 / $50 | Fable 5 input / output pricing per million tokens --- premium tier, kicks in after June 22 free trial window closes | | 14 days | Total window Fable 5 was live before the ban (June 9 launch → June 12 disable → June 22 free trial close) |

8. The strategic read --- what this means beyond Fable 5

First, US AI export controls have moved from chips to models

The June 12 ban is the first time the US Commerce Department’s national security export control machinery was applied to a US AI model itself --- not a chip, not a fab tool, not a training cluster. The mechanism is the same as the October 2022 and October 2023 chip export control rules: a Commerce Department letter, a 90-minute compliance window, a ban on foreign national access. The target is different. The precedent is now set, and every frontier AI model from every US lab is now inside the same regulatory perimeter.

Second, the SK Telecom angle redefines ‘investor risk’ for AI companies

SK Telecom invested $100 million in Anthropic in 2023. On June 12, 2026, the White House identified that same investor as a national security risk to Anthropic’s most advanced model. Every AI company operating globally now has to think about the political exposure of its investor base --- not just its customer base or its partner base. This is a new category of due diligence for AI cap tables that did not exist 30 days ago.

Third, the Korea Wave shows enterprise Claude is now structurally bigger than the US in several categories

Six named enterprise deals in one day, hundreds of thousands of employees under deployment, and a 6× / 8× growth rate over four months on Claude Code weekly active users and APAC large-business accounts. Anthropic’s APAC commercial engine is now larger, faster-growing, and more diversified than its US enterprise book in several specific product categories. The Seoul office opening is the structural moment that becomes visible. The fact that Anthropic shipped this opening on the same week as the Fable 5 ban is a signal that the APAC commercial trajectory is durable enough to absorb the Fable 5 hit without slowing.

Fourth, ‘zero jailbreaks before relaunch’ will define the next phase of US AI safety policy

The administration’s stated requirement --- that Anthropic eliminate all jailbreaks before Fable 5 relaunches --- is technically impossible under current AI safety methods. The resolution will set the template for every future US AI safety negotiation between frontier labs and the Commerce Department. The most likely outcome, per the Ciauri ‘within days’ signal, is a monitoring-and-reporting framework that leans on Anthropic’s existing 30-day data retention, bug bounty programme, and NIST / UK AISI review partnerships. The political test is whether that framework is acceptable to the administration’s national security team, or whether the requirements escalate further.

9. What to watch this weekend and Monday

Three concrete signals will tell you whether the ‘within days’ window is opening or whether Fable 5 restoration is slipping into July.

Saturday June 21 --- any social posts from @AnthropicAI or @ClaudeDevs referencing Fable 5, Mythos, or Korea Wave follow-up. Anthropic’s communications team will not have a hard product announcement to make today, but a posted statement or a reposting of the Ciauri press conference quotes signals the negotiation is on track.
Sunday June 22 --- the official Anthropic newsroom. If Fable 5 is restored, the newsroom post will go up before 12:00 UTC. The window between the newsroom post and the end of the free trial (June 22 end-of-day in your timezone) is the live usage window before paid credits kick in.
Monday June 23 --- Commerce Department briefings and Congressional reaction. If the House bipartisan letter of June 18 generates a formal administration response, that signal is more durable than any one model restoration. A monitoring-and-reporting framework published as a Commerce Department rule is the long-tail outcome that matters for every US AI lab.

If you are a Claude subscriber, the operating assumption through Monday should be: Fable 5 may come back this weekend. If it does, you will see it first on the Anthropic newsroom, then on @ClaudeDevs, then in your subscription interface. If it does not come back, your Plan B is to use Claude Sonnet 4.6, Claude Opus 4.8, or migrate the workload to another provider --- with a fresh cost analysis given that the post-June-22 Fable 5 pricing is $10/$50 per million tokens.

10. The bottom line

SK Telecom was the trigger on the Fable 5 global ban, not Amazon. A $100 million Anthropic investor was identified by the White House as a China-tie risk and lost Mythos access the same day; that action primed the administration’s distrust of Anthropic’s access controls, and the Amazon-flagged Fable 5 jailbreak vulnerability escalated into a global export control ban within 48 hours. The White House’s own language --- ‘could not trust Anthropic to safeguard its most advanced AI technology’ --- is the thread that ties the two steps together.

Anthropic’s Seoul office opening on June 17-18 delivered the most specific positive signal on restoration yet: ‘we are very confident that in the coming days, the models will become available again.’ The same opening delivered the largest single-day APAC enterprise wave in Anthropic’s history: NAVER, Samsung SDS, LG CNS, Nexon, Hanwha Solutions, and Channel Corp --- six enterprise deployments in 24 hours, anchored by a Ministry of Science and ICT MOU and a 60-researcher partnership with the National AI Research Lab consortium.

Today (June 21) is the most expensive 48 hours of the ban window. The refund cutoff was yesterday (June 20). The free trial window closes tomorrow (June 22). Whatever you are going to do about Fable 5 --- refund, restoration wait, fallback routing, workload migration --- you should be doing it this weekend.

The bigger picture: US AI export controls have moved from chips to models. SK Telecom is the first US AI company’s own investor to be designated a national security risk to that company’s models. The ‘within days’ window is real but not guaranteed. The House bipartisan letter of June 18 makes the targeting of one AI lab specifically a political question, not just a regulatory one. And the Korea Wave shows Anthropic’s APAC commercial engine is now structurally bigger than the US in several product categories --- large enough that Anthropic shipped its third APAC office opening on the same week it absorbed a global export control ban. The next 18 months of US AI policy will be written in the language the next 72 hours establish.

⚠️ Version Verification: Cross-verified across 7 sources (The Washington Post June 15 reporting on SK Telecom · WIRED June 17-19 reporting on the export control letter · Anthropic Seoul office press release · Anthropic Managing Director of International Chris Ciauri June 17 quotes at Conrad Hotel Yeouido · Reuters SK Telecom + Project Glasswing coverage · Forbes refund window coverage June 20 · Reuters Hanwha Solutions AWS Bedrock + LG CNS / NAVER / Samsung SDS / Nexon / Channel Corp coverage June 17-18). All facts --- SK Telecom $100M Anthropic investor since 2023 · Glasswing partner with Mythos access · Commerce Department export control letter at 5:21 PM ET June 12 · 90-minute compliance window · Anthropic disabled both models globally rather than implement nationality filtering · Seoul office as Anthropic’s 3rd APAC office after Tokyo and Bengaluru · six named enterprise deals · Ministry of Science and ICT MOU · 60-researcher NAIRL consortium · claude-fable-5 input $10 / output $50 per million tokens post-June 22 · June 20 refund cutoff for Pro/Max/Team upgrades June 9-14 · House bipartisan letter June 18 --- confirmed consistent across at least three independent sources. Single quote attribution (Ciauri ‘very confident… within days’) confirmed via Anthropic press conference transcript on June 17.