U.S. Court Greenlights Binance’s $2.7B Settlement With CFTCCrypto Daily

cryptodailyPublié le 2023-12-12Dernière mise à jour le 2023-12-19

Résumé

Among the stipulations imposed on Binance, the court order dictates that the exchange will no longer permit existing sub-accounts, including those opened by prime brokers, to bypass the platform’s compliance controls. Additionally, the ruling requires the company to offboard every account failing to meet proper KYC compliance requirements. 

Table of Contents

As per the terms of the settlement, former Binance CEO Changpeng “CZ” Zhao will be shelling out $150 million, while Binance will pay $2.7 billion to end the CFTC enforcement action.
Settlement Breakdown
In a significant development, the U.S. District Court for the North District of Illinois has officially approved the settlement between the Commodity Futures Trading Commission (CFTC) and cryptocurrency exchange Binance and former CEO Changpeng Zhao. 
The CFTC disclosed in a press release on December 18 that the court issued a consent order encompassing a permanent injunction, civil monetary penalty, and equitable relief against Binance and Zhao.
According to the court's ruling, Zhao is obligated to pay a hefty $150 million civil monetary penalty to the CFTC. Simultaneously, Binance is mandated to make two payments, each of $1.35 billion, to serve as a refund and penalty to the CFTC. This settlement comes in response to the court's determination that Zhao and Binance flagrantly violated the Commodity Exchange Act (CEA) and CFTC regulations.
Court Findings 
The court's findings highlight that Binance, under Zhao's direction, actively solicited customers in the United States, including quantitative trading firms. Notably, these customers engaged in digital asset derivative transactions directly on the Binance platform. 
Furthermore, the court revealed that Binance, in defiance of its own Terms of Use, permitted at least two prime brokers to establish "sub-accounts" exempt from Binance's know your customer (KYC) procedures. This enabled U.S. customers to directly trade on the platform, a clear violation of regulatory standards.
Resolutions and Acknowledgment
In a blog post dated November 21, the Binance team announced that it had been able to reach resolutions with various regulatory entities, including the CFTC, Department of Justice, Office of Foreign Assets Control, and Financial Crimes Enforcement Network. 
The post also disclosed that Zhao, Binance's founder and CEO, agreed to step down and plead guilty to violating criminal anti-money laundering requirements. The team, in a candid admission of fault, stated, 
"Binance grew at an extremely fast pace globally, in a new and evolving industry that was in the early stages of regulation, and Binance made misguided decisions along the way. Today, Binance takes responsibility for this past chapter."

Compliance Measures
Among the stipulations imposed on Binance, the court order dictates that the exchange will no longer permit existing sub-accounts, including those opened by prime brokers, to bypass the platform’s compliance controls. Additionally, the ruling requires the company to offboard every account failing to meet proper KYC compliance requirements. 
Furthermore, Binance and Zhao are obligated to certify the implementation of a corporate governance structure, including a Board of Directors with independent members, a Compliance Committee, and an Audit Committee.

Lectures associées

Il y a quelques instants, DeepSeek V4 a mis à jour DSpark, augmentant la vitesse d'inférence de 80 %

DeepSeek a annoncé une mise à jour majeure de son modèle DeepSeek-V4-Pro avec le lancement de DSpark, un nouveau cadre de décodage spéculatif (Speculative Decoding) open source, accompagné de l'infrastructure complète DeepSpec. Cette mise à jour, axée sur l'ingénierie et non sur les capacités fondamentales du modèle, vise à accélérer considérablement l'inférence des LLM en production. Le cœur de DSpark est l'introduction d'une **génération semi-autorégressive**. Cela combine la génération parallèle à haut débit d'un modèle "brouillon" léger avec une modélisation des dépendances entre tokens pour améliorer le taux d'acceptation. De plus, un système de **vérification planifiée par confiance et conscient du matériel** (Confidence-Scheduled Verification) évalue dynamiquement la probabilité que chaque token généré soit accepté par le modèle cible. Il adapte ainsi la longueur de vérification en temps réel pour optimiser l'utilisation des ressources GPU, notamment en charge élevée. Les résultats sont significatifs : face aux méthodes de pointe comme Eagle3 et DFlash, DSpark augmente la longueur moyenne de tokens acceptés de 26.7% à 30.9% et de 16.3% à 18.4% respectivement sur les modèles Qwen3. En déploiement réel, par rapport à la génération token par token précédente (MTP-1), **DSpark améliore la vitesse de génération pour l'utilisateur de 60% à 85% pour le modèle Flash et de 57% à 78% pour le modèle Pro**, à débit total constant. Le projet open source **DeepSpec** fournit une stack complète pour entraîner et évaluer des modèles brouillons pour le décodage spéculatif, supportant actuellement les algorithmes DSpark, DFlash et Eagle3, ainsi que les familles de modèles cibles Qwen3 et Gemma.

marsbitIl y a 5 h

Il y a quelques instants, DeepSeek V4 a mis à jour DSpark, augmentant la vitesse d'inférence de 80 %

marsbitIl y a 5 h

Trading

Spot
活动图片