Mandatory federal reporting trigger. Amodei proposes 10^26 FLOPs for frontier CBRN/autonomous cyber capability checks.
Independent third-party red-teaming and pre-deployment clearance before cluster weights are exposed or trained.
Whether creators of open-weight models remain liable for downstream finetuning and catastrophic weaponization.
Executive reluctance to handicap US frontier race against geopolitical rivals and avoid regulatory capture.
Governance Alignment Matrix
| Actor | Compute Trigger | Open Source | Auditing | WH Likelihood |
|---|---|---|---|---|
| Dario Amodei | 10²⁶ FLOPs | Strict on Tier 4 | Gov Red-Teams | Low (20%) |
| Sam Altman | 10²⁵ FLOPs | Tiered Exemption | IAEA-style Agency | Medium (35%) |
| Elon Musk | 10²⁴ FLOPs | Restricted Open | Mandatory Pause | Very Low (10%) |
| Active Preset | 10²⁶ FLOPs | Strict Dev | Tier 4 Clearance | 24% |
Executive Feasibility & Geopolitical Friction Analysis
The simulated policy mirrors Dario Amodei’s Anthropic essay calling for government licensing and mandatory testing above 10^26 FLOPs. While endorsed in principle by Sam Altman and Elon Musk, the White House views federal pre-clearance as an unacceptable brake on US competitiveness against China, producing a feasibility score of only 24% and an estimated $1.4B annual federal enforcement overhead.