The Trump administration reportedly ordered AI firm Anthropic to immediately revoke South Korea's largest telecom operator SK Telecom's access to its most advanced AI model, Claude Mythos, a move that later escalated into broader export controls, WIRED reported on June 17, 2026. The reason cited was concern over SK Telecom's "suspicious links" to China.
June 2026 · AI Export Controls
US Pulls SK Telecom's Access to Anthropic's Most Advanced Model Over China-Link Fears
The White House asked Anthropic to revoke SK Telecom's access to "Claude Mythos," then ordered restrictions on all foreign nationals — prompting Anthropic to take the entire model offline. SK Telecom flatly denies any ties to China.
~150
trusted orgs in Project Glasswing — SK Telecom was one
$100M
SK Telecom investment into Anthropic (2023)
OFFLINE
entire Mythos model temporarily taken down
The "China link," to scale
SK Telecom's actual China-related footprint is tiny next to its bet on Anthropic — the suspicion traces to a residual stake in a 2004 SK Group joint venture.
$1.9M China revenue7 local staff
How it unfolded
Concern raised Officials flag SK Telecom's alleged China links + Mythos access
→
Access revoked White House asks Anthropic to cut access — Anthropic complies
→
Broad restriction All foreign nationals barred; model taken fully offline
→
Negotiating Anthropic & White House weigh restoring the model
SK Telecom's response
"No factual basis behind the claims of anonymous sources" — the company says it has no ties to China . Korean reporting pushed back on the allegations as unfounded.
Wider fallout
Fears of spillover to other Korean firms — Samsung, KISA, SK hynix — amid tightening US AI export controls and US–China tech rivalry.
Why Mythos is so sensitive
Claude Mythos 5 is described as exceptionally capable at detecting cybersecurity vulnerabilities — the core fear being such capability could reach China. A safety-hardened version, Fable 5, shipped June 9, 2026, but was later flagged for a bypassable guardrail , reportedly shaking government trust in Anthropic.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…