what the fuck! Silicon Valley has recently staged a big drama of tearing apart Anthropic CEO fires first, Sam Altman, and then Palantir CEO joins the battle! That's how it happened On the same day. The Pentagon terminated its cooperation with Anthropic first. The reason is simple: The security terms are too strict. A few hours later. OpenAI announces collaboration with the Pentagon. He directly snatched the Anthropic bowl from Claude's parent company Then the super rip off in the AI community began. 1. OpenAI's' AI Security ', 80% is about making appearances Firstly, the Anthropic CEO sent an internal letter to the entire company. The core meaning is just one sentence: OpenAI's so-called AI security mechanism 20% is for protection. 80% of it is just for show. The structure of OpenAI is roughly as follows: The model allows all legitimate uses. Add another layer of 'security layer' outside. for example - Refusal mechanism - Classifier - Manual review Question is: The model has no idea where it is being used. It doesn't know: Is it in the weapon system Are you conducting large-scale surveillance Are you analyzing data on American citizens The model only sees input and output. The real context cannot be seen. 2. The key to the breakdown of negotiations: a 'data analysis' clause The final stage of negotiations. The Pentagon has proposed a condition: As long as Anthropic deletes one clause. The contract can be signed. That one says: Prohibit AI from being used for 'batch data analysis'. Simply put, it means: The government can buy GPS behavioral trajectory Consumption Record Various user data Then use AI for large-scale analysis. Anthropic thinks this is the bottom line. Refuse to delete. The negotiations broke down directly. 3. Anthropic Accusation: Sam Altman's Three Step Trick Anthropic directly named Sam Altman in the letter: the first step Publicly speaking: I also support the principle of safety. Step two Sign the contract behind the scenes and take over your position. Step three Further external narration: Anthropic is too difficult to collaborate with. Anthropic is too idealistic. Anthropic is not flexible enough. There is a term in psychology: Gaslighting (gas lamp control). 4. There are political factors behind it The Anthropic CEO was very straightforward in his letter. The reason for their tense relationship with the government is: Not donating to the Trump campaign Not publicly praising the government Support AI regulation And OpenAI co-founder Greg Brockman To Trump's Super Political Action Committee Donated 25 million US dollars. In Washington, this kind of thing carries weight. 5. Palantir CEO also went off to criticize Palantir CEO Alex Karp said: If Silicon Valley wants to take away all white-collar jobs, Not willing to support the military, The final result is only one: Technology will be taken over by the state. Anthropic also counterattacked: The AI security system sold by Palantir, Basically, it's just decoration. 6. The competition for AI talent has become so crazy that it's outrageous Anthropic CEO just said: Competitors offer their researchers a price: 100 million to 500 million US dollars. single. Not the company. As a result, many people did not leave. Because the company told them: You are here for a mission. Finally, to be honest. The core of today's confrontation is not actually safe. Not a talent either. But rather a more realistic question: Who will become the AI supplier for the national machine. military intelligence system government These are the biggest customers of AI. Something is happening in Silicon Valley right now: AI companies are beginning to integrate into the military industrial system.