An investigation by The New York Times, published on September 21, 2026 by journalist Kirsten Grind, revealed that Silicon Valley 'hacker houses' — mansions where OpenAI and Anthropic engineers and startup founders pay up to $10,000 a month for a room to gain access to the industry elite — have become sites of dangerous parties. At the most famous of them, AGI House in Hillsboro, police have recorded 37 incidents since 2022, and previously unpublished police reports and court records document complaints of sexual harassment and rape. This is the first systematic investigation of the closed social environment around AI communes, built on verifiable documents — and at the same time a specific risk map for the industry and for anyone considering living in a hacker house for networking.

image
image

What happened

The New York Times piece by Kirsten Grind was published on September 21, 2026, and relies on previously unpublished police reports and court records, not just anonymous claims. Yehong Zhu, a 30-year-old founder of AI startup Zette, filed a police report in December 2021 with a rape accusation against a former resident of the tech house Genesis; the case is still under investigation. In another episode, according to a police report, a woman had to escape from AGI House by climbing over a gate to get away from one of the residents. At least 17 of the 37 police incidents at AGI House are neighbor complaints about parties; at another house, more than 20 complaints accumulated over 2021–2024, and the damage to the landlord exceeded $200,000. Allegations of sexual harassment and rape in the investigation are documented specifically by law enforcement and court records.

Context

'Hacker houses' are a format of paid co-living where a room is rented not for housing, but for access to the industry elite: this is how part of the social infrastructure around the AI industry is set up. The most famous house, AGI House, was created by former Google Brain researcher Jeremy Nixon after a trademark dispute with Rocky Yu; among the house's guests were Sergey Brin and Grimes, and the venture fund associated with the house, according to the house's own website, was supported by Eric Schmidt and Marc Andreessen — Andreessen denies his involvement. Sam Altman, according to his representative, spoke once at the Residency hacker house network. Important background: the very model of 'pay for a room for access to the elite' is a documented example of how career capital in AI is distributed through informal social networks, so getting into such houses provides a networking advantage not directly tied to professional results.

Why this matters for the industry

For the industry, this is the first systematic investigation based on police reports and court records about the closed social environment around AI communes: the ecosystem, which includes OpenAI and Anthropic employees, now faces a specific question about codes of conduct and the responsibility of organizers of such spaces. The immediate effect is reputational and discursive: everyone who has publicly spoken or lived in these spaces may face questions from the press, investors, and candidates, and there is no measurable community reaction yet — discussion on Hacker News is at 2 points with 0 comments, meaning the signal is weak and may either spread across news cycles or fade. The story does not affect deployment, the engineering stack, or inference cost, but it exposes a lack of trust & safety in offline environments: informal spaces operate without basic participant verification, incident registration, and reputation layers, and this is an unoccupied product niche — a window for platform verification and reporting services while the category is empty. The likely scenario in six months is formalization: codes of conduct in well-known houses, internal lab guidelines for employee-residents, and re-investigations based on public records. On a two-year horizon — either institutionalization with legal entities, insurance, and incident reporting, or a loss of legitimacy and contraction: the paid access model will likely be devalued, and part of the networking will migrate to employer-controlled formats.

Why this matters for users

If you were considering living in a hacker house for networking, you now have a documented risk map instead of promises of 'access to the elite': verifiable incidents, complaints, and financial damage at specific houses, not abstract rumors. Practical takeaway for a potential participant: before paying for a room or going to a party, check whether the format has basic processes — a code of conduct and an incident reporting channel — and consciously assess who you are giving money to and what environment you will be in; organizers of 'informal' formats should ask themselves the same questions from their side. The investigation shifts the criteria for choosing networking spaces: safety and the presence of procedures become part of the price tag on par with the composition of residents and proximity to the industry 'elite'.

What is still unknown / limitations

The main caveats concern data quality. All key numbers — 37 incidents, at least 17 neighbor complaints, more than 20 complaints over 2021–2024, landlord damage over $200,000 — are absolute, without a denominator: it is unknown how many parties, residents, and guests there were over the period, so neither the frequency of risk nor its dynamics can be assessed; moreover, one count mixes trivial events (noise) and serious ones (violence allegations). Filing a police report is not proven guilt: key criminal episodes have not yet reached a court outcome. Attribution is a separate trap: visits by Sergey Brin and Grimes, the house's connection to a venture fund (according to the house's website — supported by Eric Schmidt; Marc Andreessen denies his involvement), and Sam Altman's one talk at the Residency network — these are associations, not control, and responsibility for what is happening cannot be transferred to them. Finally, the original NYT investigation from September 21, 2026 is available in open retellings by NY Post and AI Weekly, and the public signal is still weak: the reaction of the industry and community may either spread across news cycles or fade.

Sources

Author

Look at AI, editorial team