I was invited to a group viewing of the Focus Features film called The AI Doc: Or How I Became an Apocaloptimist. It was released March 2026, as a documentary by Daniel Roher and Charlie Tyrell built on interviews with the heads of OpenAI, Anthropic and Google DeepMind.
For someone formally trained as a historian who has been on the forefront of AI security research for over 15 years, I found it difficult to watch. For example, partway through, a famous voice in the film reaches for a particular standard analogy. He says Germans were working on the bomb in the 1940s and…
so it’s not so easy to tell Robert Oppenheimer then, “slow down.”
Record scratch. I almost walked out of the film at that moment.
The line sounds like history. It is NOT. That is propaganda. Mythology at best.
On every point I can refute it. The German threat was a belief, the belief was exposed as such and thus failed before the bomb existed. People told Oppenheimer to slow down, in writing, and he was the one who refused to listen.
It’s a FACT that Oppenheimer was told to slow down. And this AI “documentary” tries to push total propaganda that Oppenheimer had no choice but to speed up?
The failure of the film is simple to explain. It takes people who are purported experts in one area, and then pushes them into a discussion outside their expertise.
At one point a man presented as some kind of AI theorist in ethics is allowed to go on-and-on about military tactics where he clearly has zero clue and should be shut down. Imagine an artist inviting an accountant to sit in a hot-seat under intense lights to explain how brain surgery works.
Let’s dig in deeper to this Oppenheimer mistake, because it’s so easily disproved. It shows the kind of fact checking devoid from the film. The lack of integrity and quality that should have gated its release.
Albert Einstein signed the famous letter to Roosevelt of 2 August 1939 because he believed Germany was ahead. His belief was mistaken. The letter warns that Berlin had stopped uranium sales from the Czech mines it had seized and that German physicists were repeating the American experiments. That belief is the entire basis of the letter, meaning we now know conclusively it wasn’t based in fact. Fifteen years later Linus Pauling visited Einstein at Princeton and wrote down what he heard. Einstein called the letter the “one great mistake in my life“, with the single excuse that the Germans theoretically could have built a weapon first. It was anxiety that drove him to press a fear and he regretted it once the fear disproved.
Oppenheimer himself rated that fear low. In a 1965 interview he said that when he went to New Mexico he worried more about the campaigns in Africa and Russia than about a German bomb, and that belief in the German threat varied from person to person. Wigner kept it until there were no Germans left.
Today there’s no doubt about the Germans falling out of any bomb race by the summer of 1942. Albert Speer met the physicists at the Harnack House in Berlin on 4 June 1942 and Heisenberg told him a bomb was beyond reach within the war. To be fair, the Nazis had lost the war by 1943, meaning the later years are best described as German genocide and suicide while refusing to surrender a lost war. The project in this context was reduced to reactor research. Heisenberg put it plainly to Der Spiegel in 1967:
There never was a race for the atomic bomb.
Three years before Trinity, the race didn’t exist. Put THAT in your AI “documentary”.
The Allies found this out while the bomb was still eight months from completion. Samuel Goudsmit, scientific head of the Alsos intelligence mission, wrote that at the fall of Strasbourg in November 1944 his team found “irrefutable documentary evidence that Germany did not have atomic weapons” and was far behind on reactors. The same month James Chadwick told Joseph Rotblat at Los Alamos that intelligence showed the Germans were off the bomb. Rotblat asked to leave at once and was home before Christmas. He had already heard General Groves say at a dinner in March 1944 that “the real purpose in making the bomb was to subdue the Soviets”.
One scientist did what the AI “documentary” calls unthinkable. He told the bomb project to stop, and he stopped.
The others told it to slow down. Do you see yet how far from documenting actual technology this “documentary” is?
The Interim Committee met on 31 May 1945, three weeks after the German surrender, and settled on use against a Japanese city without warning. The scientists at the Chicago laboratory replied within days. The Franck Report of 11 June 1945 asked for a demonstration before any use on a city and warned that a surprise attack would start an arms race. Gee, that seems an important detail for an AI film. Leo Szilard’s petition of 17 July 1945 carried 70 signatures and asked the President to withhold the bomb until Japan had been given terms and refused them.
Oppenheimer is the guy who blocked the petition at Los Alamos. He refused to be told to stop or slow down.
Szilard sent it to Edward Teller to circulate there. Teller recalled that Oppenheimer talked him out of it on the grounds that scientists had no business applying political pressure, and the petition collected no signatures at the laboratory. Oppenheimer had already signed, for the Scientific Panel, the recommendation of 16 June 1945 that disposed of the demonstration idea:
we see no acceptable alternative to direct military use.
Then it was used, twice, by the military on a country that had nothing to do with the bomb development. The order of 25 July 1945 authorised the first bomb after 3 August and added that “additional bombs will be delivered on the above targets as soon as made ready”. The second strike needed no further decision. Hiroshima was 6 August. The Soviet Union declared war on Japan on 8 August. Nagasaki was 9 August.
The steps are important, especially since Americans never learn this correctly. The Japanese Supreme War Council was called on the morning of 9 August in response to the Soviet attack, before the second bomb fell. The Soviet entry weighed at least as heavily as the bombs on the surrender decision, and the bombs added little to what conventional raids had already done: Tokyo, Kobe and Yokohama were each more than half burned out by August 1945. The atomic bomb was a demonstration aimed at the Soviets, not so much at the Japanese.
After the Soviet test of August 1949 the question became the hydrogen bomb, and Oppenheimer, chairing the General Advisory Committee, finally gave the advice he had refused in 1945. The committee’s report of 30 October 1949 recommended against a crash programme, and the majority annex he signed called the weapon one of genocide. Bit late, no?
Truman overruled him on 31 January 1950. Four years later the Atomic Energy Commission stripped his clearance. In December 2022 the Department of Energy vacated that decision and found the 1954 process had been driven by a wish to discredit him in the public debate over nuclear weapons policy.
When Oppenheimer finally said slow down in a way that counted, the time he waded into politics as he had told others not to, the state ended his career for it. Think about that political reason for not saying slow down, versus the actual risks from technology.
So the AI “documentary” has its technology history exactly backwards.
Telling Oppenheimer to slow down was easy and it was done, repeatedly. Done by Franck, by Szilard and 69 colleagues, by Rotblat with his feet, and in the end by Oppenheimer himself. The German justification was a belief dated to 1939, which had to be abandoned in Berlin in 1942, and was disproved in Strasbourg by 1944. The project ran straight through all three dates to a different objective entirely. That is the Oppenheimer story told properly.
When an AI executive says the big scary adversary makes slowing down big scary technology impossible, the historical question is what the project did on the day the bear turned out to be a chihuahua (reactor experiment in a cave).