Breaking

Anthropic in Daily Talks With Trump Administration as Top Models Stay Paused

June 18, 2026 at 09:50 EDT

Policy & Regulation
Foundation Models
Security

Anthropic co-founder Jack Clark said the company is holding daily discussions with the Trump administration to resolve security concerns over its most advanced models, Claude Fable 5 and Claude Mythos 5, both of which have been suspended for all users following export controls.

June 2026 · Anthropic

Too Dangerous to Ship: Anthropic's Frontier Models Meet the White House

Anthropic is in daily talks with the Trump administration over its two most capable AI models. After their June 9 debut, export controls landed and the top-tier Mythos 5 was pulled — its vulnerability-discovery powers deemed too risky for public release.

Daily

Talks with the administration to resolve security concerns

Frontier models at the center: Mythos 5 & Fable 5

91/100

Senior-engineer benchmark vs 63 (Opus 4.8), 62 (GPT-5.5)

FrontierCode (Diamond) — coding at the frontier

Share of hardest coding tasks solved. Heights drawn to scale.

29.3%

Fable 5

13.4%

Opus 4.8

5.7%

GPT-5.5

SWE-Bench Pro: Fable 5 80.3% vs Opus 4.8 69.2%

Claude Mythos 5 — locked down

Top tier. Standout vulnerability-discovery; "too dangerous" to release publicly. Select partners only (e.g. Project Glasswing).

Claude Fable 5 — released with guardrails

Mythos-class power plus safeguards blocking biology & cyber. General availability for enterprise & paid users.

Pricing — per million tokens

$10

Input

$50

Output

~2× the price of Opus 4.8 — but less than half the cost of the Mythos Preview.

▲ What developers praise

Handles complex tasks over hours
Strong end-to-end software development
Builds agents & manages long-running projects

▼ The criticism

High latency and high cost
Less suited to time-sensitive uses
Not ideal for beginners

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.