
Summary of "some" of Shivon Zilis’s Betrayals of Elon Musk
This comes from the OpenAI trial and consists of proven evidence; the rest is fantasy.
1. Real-Time Intelligence Leaks and Espionage
- PX-167 (February 17, 2018): While Musk trusted Zilis to handle communications, she acted as an informant for Sam Altman. Just a day after telling Musk she would stay "close and friendly" to OpenAI to "keep info flowing," she was feeding Altman exact details regarding Musk’s recruitment efforts for Tesla, filtering out candidate names like Wojciech and Scott, and probing whether Musk had reached out to the DOTA team.
- "Trust Frameworks": In the exact same exchange, right after leaking Musk’s moves, Zilis asked Altman to establish secret "trust frameworks" so they could continue collaborating and assisting each other behind Musk's back without getting caught.
2. Gradual Normalization of the For-Profit Model
- DX-0627 (July 18, 2017): Zilis casually slipped the idea of converting OpenAI into a for-profit entity ("assuming OpenAI converts from being a non-profit") to Musk and Sam Teller. She presented the abandonment of the original non-profit mission as a minor, technical premise needed for a hardware deal (Cerebras), gradually eroding Musk’s resistance through repetition.
- Shielding Behind Others & Weaponizing Paranoia: In DX-0627, she used Greg Brockman and Ilya Sutskever as shields to push the narrative that AGI required "prioritizing speed over safety" to beat Google/DeepMind. She continuously fed Musk’s concerns about Demis Hassabis to validate abandoning the original open, non-profit ethos.
3. Architecting the "Divestment Blueprint" & Programmed Dilution
- DX-0664 (August 30, 2017 – OpenAI Starting Sketch): Zilis drafted a governance and financial structure designed as a trap. On paper, she proposed that Musk provide $110 million (81% of initial cash) for a 55% stake. In the fine print, however, she explicitly outlined a plan to reduce him to a minority stake in just 1–2 funding rounds, creating the legal machinery to strip him of control once his capital was secured.
- Nullifying Veto Power (Creative Overrule): In both DX-0664 and DX-0657, she proposed an expanded board of directors with a unanimous "overrule provision," ensuring Musk would be permanently blocked by a unified front aligned with Altman and Brockman.
4. Actively Negotiating Against Her Own Employer
- DX-0657 (August 27, 2017): In an email to Jared Birchall and Sam Teller following a two-hour meeting with Brockman and Ilya, Zilis acted as OpenAI’s spokesperson rather than Musk’s advocate. She framed it as an "absolute non-negotiable" that Musk (or anyone else) have total control over AGI.
- The "Time Trap": In DX-0657 (and her subsequent email to Musk on August 28), she tied executive control to physical time commitments (requiring 5–10 hours per week). Knowing Musk was running Tesla and SpaceX, this was a calculated trap designed to create a "logical" excuse to strip him of operational command.
- Advocating Equity for OpenAI Founders: Zilis pushed for substantial equity stakes for OpenAI's founders and an expanded employee pool, actively structuring the commercialization of the company against Musk's interests.
5. Covert Backdating and Covert Information Brokering
- Backdated Resignation: On March 25, 2018, Zilis asked Musk to send a resignation note to the OpenAI board backdated to February 20, 2018, helping clean up the timeline while she was already fully operational in alignment with Altman.
- Using Intermediaries for Psychological Manipulation: She consistently routed messages, drafts, and analytical breakdowns through Sam Teller and Jared Birchall to manipulate Musk indirectly and soften his stance before approaching him directly.