Last updated:
Hefu Chai (born c. 1990) is a software engineer based in Menlo Park, California. On September 15, 2026, federal prosecutors charged Chai, then an OpenAI engineer and formerly a senior staff engineer at Robinhood, with commodities fraud and wire fraud. Prosecutors allege that while at Robinhood he misappropriated confidential information about upcoming cryptocurrency listings and used it on at least ten occasions between 2025 and January 2026 to take long positions in Hyperliquid perpetual futures shortly before public listing announcements, generating more than $50,000 in alleged illicit profits.[1][2][7][11]
Chai completed his undergraduate studies at Shanghai Jiao Tong University in China, where he earned a Bachelor's degree with a major in Computer Software Engineering and a minor in Finance. During his time there he received several academic distinctions, including the National Scholarship of China awarded by the Chinese Education Ministry in November 2011, in which he ranked 4th out of 170, and two Scholarships of Academic Excellence from Shanghai Jiao Tong University, awarded in November 2010 and November 2012.[3]
He subsequently pursued a Master's degree at the School of Computer Science at Carnegie Mellon University in Pittsburgh, Pennsylvania. His graduate coursework concentrated on data systems, distributed computing, and machine learning, spanning subjects such as advanced storage systems, database systems, parallel computer architecture, and multimedia databases and data mining.[3]
Chai’s LinkedIn profile lists systems and machine-learning projects in connection with his Carnegie Mellon studies. These include a simplified MapReduce framework and distributed file system inspired by Hadoop, supporting file replication, fault tolerance, and failure recovery, and a hybrid storage system using the FUSE toolkit that combined SSD performance with cloud-storage capacity through segment-level deduplication, compression, and caching. His GPU-computing projects included accelerating backpropagation through time for training LSTM neural networks using the CUDA Thrust and CUBLAS libraries; he reported a 20-fold speedup over a Matlab CPU implementation. Other listed projects covered graph mining across big-data systems such as HP Vertica, Hadoop, Apache Hama, SciDB, and PostgreSQL; rewriting graph algorithms including PageRank, triangle counting, and belief propagation in plain SQL; and research on load balancing of the Bulk Synchronous Parallel model in multi-core environments.[3]
A presentation titled "Invisible loading: Access-Driven Data Transfer from Raw Files into Database Systems" names Chai as its presenter and examines how to reduce the "time-to-first-analysis" for large scientific and social-network datasets.[4] The presentation describes a technique for incrementally loading data into a database system as a side effect of running MapReduce jobs over a distributed file system, with the goal of combining Hadoop's low startup cost with the long-term query performance of a database. Building on ideas from HadoopDB and using components such as HDFS and MonetDB, the design described in the presentation loaded only the attributes actually processed by a job, used address columns to track tuple movement during sorting, and employed an "incremental merge sort" approach intended to keep query effort bounded. The presentation reports that the method imposed almost no burden on MapReduce jobs and offered relatively low cumulative cost, while noting weaknesses such as data duplication without garbage collection and suitability mainly for short-lived data.[4]
Chai's early roles were brief. He worked as a software engineer at Morgan Stanley in Shanghai in 2012, in Technology and Information Risk, and later as a software engineer at Google in Mountain View, California, where he worked on the AdWords serving infrastructure within its advertising business.[3]
From January 2015 to January 2021, Chai was a Staff Software Engineer at Facebook, based in Menlo Park, California, working on Instagram Infrastructure and Messenger. According to his own account, he led and built a self-serve platform for demand and efficiency across web, machine-learning, caching, and storage systems. On Messenger he led development of the product's first version of ephemeral messaging features and its in-app games, contributed to modernizing the Messenger camera for augmented reality, and helped enable branded AR effects.[3]
Chai joined the brokerage firm Robinhood in January 2021 as a Senior Staff Software Engineer and remained there until May 2026, according to court filings and local reporting.[3][7] He states that he led the development and launch of several of the company's cryptocurrency products, including perpetual futures, crypto staking, and Robinhood Connect, a crypto on
In June 2026, after leaving Robinhood, Chai joined OpenAI as a Member of Technical Staff, based in the San Francisco Bay Area and affiliated with the company's Codex product, an AI coding tool.[3] Working on the openai/codex repository, he is reported to have developed a backend feature enabling real-time querying of the pairing status of remote control devices, implemented as a status transport layer written in Rust and TypeScript with an experimental gRPC-based endpoint. The feature allowed the desktop application to poll pairing status using either a pairing code or manual entry, establishing an end-to-end status flow between the backend and the desktop app that is reported to have improved onboarding reliability.[5]
On Tuesday, September 15, 2026, the U.S. Department of Justice announced charges against Chai and a co-defendant, Huaisong Xiang, a 30-year-old Robinhood engineer from Jersey City, New Jersey. Each was charged with one count of commodities fraud under the Commodity Exchange Act, which carries a maximum sentence of ten years in prison, and one count of wire fraud, which carries a maximum sentence of twenty years.[1][2][11]
Prosecutors allege that from March 2025 through February 2026, Chai and Xiang had access to nonpublic information about whether and when Robinhood Crypto planned to add particular cryptocurrencies to its trading platform, and that they repeatedly purchased perpetual futures tied to those tokens on Hyperliquid, a decentralized perpetual futures trading platform, before Robinhood publicly announced the listings.[7][8] According to the criminal complaints, Chai allegedly routed his trading through three Hyperliquid wallets that investigators linked by tracing transfers to and from an exchange account he opened in December 2017, and on at least ten occasions in 2025 and January 2026 bought long positions in perpetual futures on tokens that Robinhood was about to list.[11][12] The government alleges that each man earned more than $50,000 in illicit profits from the trades, while at least one account summarizing Department of Justice materials describes the alleged aggregate illicit profits as approximately $1.13 million.[1][2][8][9] The complaint against Chai lists alleged trading episodes tied to tokens including MEW, MOODENG, ASTER, XPL, HYPE, ENA, AERO, SYRUP, LDO, DOT, and LIT.[12] The complaint alleges trades in perpetual futures, which are derivatives, rather than in the underlying tokens. This differs from a 2022 Justice Department prosecution over advance knowledge of Coinbase token listings, which the department called the first cryptocurrency insider-trading tipping scheme.[1]
The criminal complaints describe both defendants as "Coin Aware Individuals," a designation that gave them access to a private Slack channel containing confidential information about planned token listings and was subject to policies barring such employees from trading affected tokens before, and for 24 hours after, Robinhood's public announcements.[8] In Chai's case, the commodities-fraud charge alleges that, knowing he had obtained material nonpublic information in breach of a duty to Robinhood, he used that information to trade perpetual futures on a decentralized derivatives exchange for personal gain, in violation of Section 6(c)(1) of the Commodity Exchange Act and Commodity Futures Trading Commission Regulation 180.1.[11] The wire-fraud count alleges a parallel scheme to defraud using interstate wire communications.[11]
Jamie McDonald, the U.S. Attorney for the Southern District of New York, stated: "Misappropriating confidential information to trade in the derivatives markets for personal benefit is illegal," adding that the charges make clear that corporate insiders cannot evade the securities and commodities laws by trading based on misappropriated information in derivatives such as perpetual futures, tokenized securities, or other similar financial instruments.[2][10] The case was investigated by the FBI and is being prosecuted by the Securities and Commodities Fraud Task Force of the U.S. Attorney's Office for the Southern District of New York.[2]
Chai was arrested on September 15, 2026 and was scheduled to be arraigned the following day in federal court in the Northern District of California, while Xiang was arrested and scheduled to be arraigned in Manhattan federal court.[7]
Commentators have described the case as an early test of applying commodities-fraud theories and insider-trading concepts to activity on decentralized derivatives platforms, noting that it follows Commodity Futures Trading Commission guidance that misappropriation of confidential information in breach of a pre-existing duty can violate Section 6(c)(1) and Regulation 180.1 in derivatives markets.[12] The charges are accusations, and both defendants are presumed innocent unless and until proven guilty.[2][6]
The November 16, 2026 CoinDesk report said the Justice Department’s announcement of Xiang’s plea did not provide an update on Chai’s case, and the report did not report a plea by Chai.[13]
On November 16, 2026, Xiang pleaded guilty to wire fraud before U.S. District Judge Colleen McMahon in Manhattan. The report said he used confidential information about upcoming Robinhood cryptocurrency listings to trade token perpetual futures on Hyperliquid, and agreed to forfeit $125,523 in profits.[13]