Palantir Co-founder Joe Lonsdale: Some Leading AI Companies Use Survival Risk Warnings to Influence Public Policy, Which is Dangerous Manipulation
Palantir co-founder and 8VC partner Joe Lonsdale, an investor in Anthropic, stated at the Reuters Momentum AI event in Austin that some leading AI companies use survival risk warnings to influence public policy, which he considers dangerous manipulation. He specifically mentioned that individuals around OpenAI and Anthropic are funding external groups to push for stricter regulations, believing this would allow a few companies to participate in writing the rules and consolidate their market position.
He remarked that the products being pushed by these two companies are very dangerous and warned against the emergence of oligarchs controlling government policy; while safety is real, fear and regulation must be responded to carefully. In the same month, Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and Elon Musk all called for leading labs to slow down the pace of developing the most powerful models, citing that safety measures are lagging behind and could pose survival risks. Lonsdale also praised Amodei, stating that Anthropic is ahead of its competitors in talent and expressed pride in investing in the company. He also mentioned that companies he supports might acquire Ukrainian defense firms.
The buyers of the fear narrative are labs that require licenses and safety budgets, along with the advocacy organizations funded by them, while the sellers are investors and defense software companies seeking open weights and fewer licenses. Funding flows from model companies into policy groups and then back into regulatory texts. The beneficiaries are those in power who can write safety into entry barriers; the pressured are the latecomer training teams that cannot understand survival risks and only see the licensing wall. The driving force behind these events is investors publicly undermining others, rather than new bill votes.
Source: Public Information
ABAB AI Insight
Lonsdale is on the shareholder list of Anthropic while also selling Palantir-style government software, meaning safety and licensing are not on the same balance sheet for him. Survival risk elevates regulation from product liability to a civilizational level, with external groups writing the elevated agenda into legislative memos, and the incumbent labs being most capable of submitting compliance materials. He acknowledges that safety is real, and separating fear is meant to preserve action space for defense and open source.
The capital path is initiative funding. Funds close to labs flow into external groups, which produce slowdowns and licensing suggestions, raising compliance costs for latecomers. If the 8VC portfolio buys Ukrainian defense assets, it needs systems that are exportable and deployable, not a global pause on the most powerful models. Praising talent while criticizing policy separates company operations from the Washington narrative.
In comparison to nuclear licenses, the Basel Accords for large banks, and FDA regulations for pharmaceuticals: safety discourse always raises fixed costs, which protect those who have already passed. AI is at a stage of "whether to turn training itself into a licensed activity," with survival risk being the highest multiplier adjective.
Structural changes belong to regulatory changes. Pricing power shifts from open-source releases and computing power purchases to who funds the rule-writers. The mechanism is: the greater the risk, the more the rules resemble entry barriers; the more entry resembles entry, the more stable the oligarchs become. Investors undermining others fear that the second place they invest in will also be written into the licenses by the first place.