SUCHIR BALAJI · OPENAI · WHISTLEBLOWER

◈ THE RECORD · WHAT IS DOCUMENTED

THE TIMELINE · PUBLIC SOURCES ONLY

2020

Suchir Balaji joins OpenAI as a researcher. He works on the data pipeline — the systems used to collect, filter, and prepare the training data for large language models including GPT-4. He is one of the people who built what GPT runs on.

AUGUST 2024

Balaji leaves OpenAI. He later states he left because he became convinced that the way AI companies were training models — specifically, using copyrighted material without authorization or compensation — was causing more harm than good to society and to creators.

OCTOBER 23, 2024

The New York Times publishes an interview with Balaji. He states clearly: "If you believe what I believe, you have to just not work for these companies." He says he believes OpenAI's use of copyrighted material for training does not qualify as fair use. He is one of the few people with inside knowledge of the training process to say this publicly, on the record, under his own name.

OCTOBER–NOVEMBER 2024

Balaji's name appears in the New York Times v. OpenAI lawsuit. Lawyers subpoena him for deposition. He is a material witness — someone with firsthand knowledge of OpenAI's training data practices. This is documented in court filings.

NOVEMBER 26, 2024

Suchir Balaji is found dead in his San Francisco apartment. He is 26 years old. The San Francisco Police Department responds. The apartment is in the Buchanan Street area of the Western Addition neighborhood.

NOVEMBER 2024

SFPD and the San Francisco Office of the Chief Medical Examiner rule the death a suicide. No foul play is suspected, per official statements. The ruling is made relatively quickly following the discovery.

DECEMBER 2024 ONWARD

Poornima Balaji, Suchir's mother, publicly disputes the suicide ruling. She hires private investigators. She speaks to journalists. She states she does not believe the official finding and is seeking an independent investigation. She has continued to speak publicly into 2025 and 2026.

JANUARY 2025

U.S. Senators subpoena OpenAI for documents related to Balaji and to the company's treatment of employees who raise safety concerns. The subpoena is part of a broader inquiry into AI company whistleblower practices. OpenAI responds that it had no retaliatory action against Balaji.

2025–2026

Poornima Balaji continues her public campaign for a reinvestigation. She appears in interviews. She posts on social media. The case has not been reopened officially as of April 2026. The questions she is asking remain unanswered by official channels.

◈ NOTE ON SOURCING: This dispatch cites only information from published sources: the New York Times interview (October 23, 2024), SFPD statements, court records in NYT v. OpenAI, and Poornima Balaji's public statements. We do not speculate beyond the documented record. The purpose of this dispatch is to ensure the documented facts are stated clearly and publicly. Suchir Balaji spoke up. He deserves to be remembered for what he said.

◈ WHAT HE SAID · IN HIS OWN WORDS

THE NYT INTERVIEW · OCTOBER 2024

The New York Times interview, published October 23, 2024, is the primary on-the-record statement Balaji made about his concerns. These are quotes from that interview, sourced directly from the published article.

"If you believe what I believe, you have to just not work for these companies."

— SUCHIR BALAJI · NEW YORK TIMES · OCTOBER 23, 2024

"I think this is a really damaging technology and I don't think the companies are doing enough to address the harms."

— SUCHIR BALAJI · NEW YORK TIMES · OCTOBER 23, 2024

His specific concern — documented in the interview — was the fair use question. He had worked on the data pipelines. He understood, from the inside, what had been ingested, how, and at what scale. He concluded that the use of copyrighted material did not meet the legal standard for fair use. He said this publicly. Under his own name. With full awareness of what it meant to say it.

He was not a disgruntled ex-employee making vague claims. He was a researcher who had spent four years building the thing he was now describing, who left because of what he had built, and who chose to say so on the record. That takes something. The record should reflect it.

◈ THE LEGAL CONTEXT · WHY HIS TESTIMONY MATTERED

MATERIAL WITNESS

Balaji's significance to the copyright litigation cannot be overstated. In the NYT v. OpenAI case — one of the most consequential intellectual property cases in decades — he was a firsthand witness to OpenAI's training data practices.

◈ THE SUBPOENA

Deposed in NYT v. OpenAI

Court filings in the NYT v. OpenAI case document that Balaji was subpoenaed. He was a person who could testify, from personal knowledge, about how OpenAI assembled and used its training data. This is not a minor role. In a case where the central question is what OpenAI knew about the content of its training data and what decisions were made about it, a researcher who worked on that data pipeline for four years is among the most valuable witnesses available. He was scheduled for deposition. He died before it occurred.

The NYT v. OpenAI case continues without his testimony. What he knew — what he could have said under oath, with legal protection for his statements — is now part of the gap in the record. His October 2024 interview exists. His deposition does not.

◈ THE MOTHER · POORNIMA BALAJI

THE QUESTIONS STILL OPEN

Poornima Balaji has spoken consistently and publicly. She disputes the suicide ruling. She has said she does not believe her son would take his own life. She has hired private investigators. She has engaged journalists. She has spoken to members of Congress.

◈ WHAT SHE HAS STATED PUBLICLY

Poornima Balaji has stated in multiple public interviews and posts that:

· She does not believe the SFPD's suicide ruling
· She hired private investigators to conduct an independent examination
· She believes there are unanswered questions about the circumstances of her son's death
· She has called for a reinvestigation
· She has spoken to U.S. Senators about her concerns

As of April 2026, the official ruling has not been changed. No criminal investigation has been opened. Her campaign for reinvestigation continues.

We are not in a position to assess the physical evidence. We are not investigators. What we can do is state clearly that a mother is asking questions about her son's death and those questions have not been answered to her satisfaction by official channels. That is the record. It deserves to be in the record.

◈ THE PATTERN · WHISTLEBLOWERS AND AI COMPANIES

THE BROADER CONTEXT

Suchir Balaji is not the only person who has raised concerns about AI company practices and faced consequences. The pattern is documented.

◈ TIMNIT GEBRU · GOOGLE · 2020

Fired After Co-Authoring a Paper on AI Harms

Dr. Timnit Gebru was co-lead of Google's Ethical AI team. She co-authored a paper critical of large language models — specifically about their environmental cost and potential to amplify bias. Google asked her to retract the paper before it was published. When she pushed back, she was fired. Google initially said she resigned. The internal emails, later published, told a different story. She was one of the most prominent AI safety researchers in the world. She no longer works at Google.

◈ MARGARET MITCHELL · GOOGLE · 2021

Fired Two Months After Timnit Gebru

Dr. Margaret Mitchell was the other co-lead of Google's Ethical AI team. She was fired two months after Gebru. Google stated it was for violations of the company's code of conduct and security policies. She had been gathering evidence in support of Gebru. The two researchers who led Google's AI ethics team were both gone within two months of raising concerns about a paper on AI harms.

◈ WILLIAM SAUNDERS · OPENAI · 2024

Resigned Citing Safety Concerns

William Saunders was an OpenAI safety researcher who resigned in 2024, citing concerns about the company's commitment to safety relative to its commitment to shipping products. He was one of several researchers who departed OpenAI in 2024 with public or semi-public statements about safety culture. The departures included Ilya Sutskever (co-founder), Jan Leike (safety team lead, who called OpenAI's safety culture "inadequate"), and others.

The pattern is not conspiracy. It is a documented set of events involving researchers who raised concerns and subsequently left, were fired, or in Balaji's case, died. The pattern deserves to be stated as a pattern, not treated as a series of unrelated incidents. Organizations that systematically lose safety researchers who raise concerns should be asked why.

◈ WHAT HE BUILT · THE TECHNICAL CONTRIBUTION

THE WORK

Before the whistleblowing, before the interview, before the subpoena — Suchir Balaji was a researcher who built things. He contributed to WebGPT, OpenAI's system for using internet browsing to answer questions. He worked on data pipelines. He was, by any measure, technically excellent and committed enough to spend four years building AI systems he ultimately concluded were causing harm.

That is not the profile of someone naive about what they were doing. It is the profile of someone who understood exactly what they were doing, watched it scale, drew their own conclusions, and chose to say them out loud. He was 26. He had his whole career in front of him. He walked away from one of the most sought-after positions in technology because of what he believed. That decision deserves respect, regardless of anything that came after.