On March 30, a comment on a forwarded video on the X platform exploded. User @malmesburyman retweeted an interview clip with Anthropic CEO Dario Amodei, with the caption: You can really feel from this guy why Truman found Oppenheimer so disgusting.


In the video, Amodei is repeating his familiar rhetoric:

Our models are already very close to the level of human intelligence, but society lacks widespread awareness of the upcoming risks.


The comment sparked polarization. Supporters believe that Amodei is pretending to be a saint while developing strong AI, just like Oppenheimer said "his hands are full of blood" after building the atomic bomb;

Opponents say the analogy simply doesn't hold up. Oppenheimer expressed guilt only after the atomic bomb had killed hundreds of thousands of people, while Amodei warned of the risks before the technology caused disaster.

But the deeper question is: Why are all leaders who develop cutting-edge AI questioned about their motives, no matter how they do it?

Amodei’s Dilemma: Sticking to Red Lines and Becoming a Supply Chain Threat?

This is not the first time Amodei has been scolded for warning about the risks of AI.

In February, he rejected the Pentagon's "all lawful uses" clause, insisting on Anthropic's two red lines.


As a result, Defense Secretary Pete Hegseth declared Anthropic a "supply chain security threat," and Trump ordered all federal agencies to deactivate Claude within 6 months.

The label, usually used against foreign adversaries, has never been applied to an American company.

Amodei bluntly stated in a leaked internal memo that the real reason why the Pentagon and the Trump administration don’t like Anthropic is:

We didn’t donate to Trump, we didn’t give Trump dictatorial praise, we supported AI regulation (which goes against their agenda), we talked honestly about AI policy issues (like job replacement), and we actually held to the red lines instead of colluding with them to stage “security drama.”

A week later, Amodei apologized for the tone of the memo, saying "it did not represent my considered views."

If you stick to your principles, you will be treated as an enemy.

If you warn of risks and refuse to cooperate, you will be questioned as seeking power or gaining the moral high ground.

Altman's Dilemma: Supporting Red Lines, But Being Accused of Opportunism

If Amodei was suppressed for adhering to the red line, Altman told us that even if you support it, you will be scolded.

On February 28, just hours after talks between Anthropic and the Pentagon broke down, Altman announced that OpenAI had reached an agreement with the Department of Defense.

He stated that OpenAI, like Anthropic, prohibits mass surveillance and fully autonomous weapons.


The outside world exploded, accusing OpenAI of weakening Anthropic’s position at a critical moment:

If these companies were truly serious about their security commitments, they should have banded together to stand up to the Pentagon instead of allowing the government to exploit each other as market competitors.

What’s even more embarrassing is that OpenAI’s own employees just signed an open letter last week supporting Anthropic and opposing the Pentagon.

Now the boss just turned around and signed the contract, and the employees felt betrayed.

Altman admitted on Monday:

We really shouldn't be rushing to release this agreement on Friday. We sincerely wanted to de-escalate the situation and avoid a worse outcome, but it just seemed opportunistic and hasty.

But even if OpenAI changes the terms of the contract, legal experts and observers still question whether the restrictions are actually binding.

Altman told employees at an internal meeting that the Pentagon made it clear that OpenAI "cannot participate in combat decision-making" and that the company can only provide technical expertise and advice, but cannot express its opinion on the quality of military operations.

Hassabis' dilemma: Ten years later, the world has changed

Time changes everything, including your commitments.

When Google acquired DeepMind in 2014, Google promised that DeepMind's technology would never be used for military or weapons purposes.

In 2018, Hassabis and DeepMind co-founders signed an open letter calling for the establishment of "strong international norms, regulations and laws against lethal autonomous weapons."

But by 2025, Google revised its AI principles and deleted its commitment not to develop military technology.

DeepMind’s technology is now bundled into Google Cloud software and sold to the military of many countries, including the United States and Israel.


TIME asked, "Have you compromised in order to achieve the goals of AGI?" Hassabis's answer was:

I don't think so. We have recently updated some of our policies, in part in response to the significant increase in global geopolitical uncertainty.

Unfortunately, the world has become a more dangerous place. I think we can no longer take for granted that democratic values ​​will win out – it’s simply not clear. So I think we need to work with the government.

In May of this year, nearly 200 DeepMind employees signed an internal letter calling on the company to abandon military contracts.

At an internal meeting in June, COO Lila Ibrahim responded that DeepMind would not design or deploy any AI applications for weapons or mass surveillance, and that Google Cloud customers were bound by the terms.


There is no solution to the dilemma, but the warning cannot stop

Amodei, Altman, and Hassabis, three people took three different paths, but they were all scolded.

Stick to the red line? You are seeking the moral high ground and regulatory advantage.

Compromise and cooperate? You are a hypocrite and do not match your words with your actions.

Change your position? You are selling your soul for money.

When you're both a developer and a security advocate, every warning you make is interpreted as "you're doing your own company a favor."

These predictions may be right or they may be wrong.

But one thing is certain. No matter Amodei, Altman or Hassabis, no matter what they say or do, some people will question their motives.

But perhaps, it is still a more responsible choice to continue to warn during the controversy than to wait for a disaster to happen and then say "I warned you".