A Souvenir from Heaven
On the day that Moses reached heaven to receive the Torah, the angels were not pleased. In fact, they did everything possible to stand in his way. But Moses was determined. And he had help from above.
How exactly did Moses reach the heavens? First he climbed to the summit of Mt. Sinai. Not bad for an 80-year old, but then what? The Midrash says that a cloud approached him and opened like a lift door; he entered the chamber, and it took him to heaven. This is how the sages understand the biblical verse, ויבא משה בתוך הענן, “Moses came within the cloud.” But when he emerged from his cloud-lift, the way was blocked by a terrifying angel named Kemuel and a force of 12,000 celestial guards.
Kemuel said to Moses, “What are you, an impure mortal, doing in this pure place of fire?” Moses boldly replied, משה בן עמרם אני שבאתי לקבל תורה לישראל, “I am Moses, son of Amram, and I have come to receive Torah for Israel.” Kemuel wouldn’t let him pass, so what did Moshe do? הכהו He punched the angel, and knocked him off the cloud.
Moses continued unhindered in heaven until he reached an even greater angel called Hadraniel who had lightning bolts shooting from his eyes. Now Moses started to weep and wanted to dive back to earth, but God had mercy and told Hadraniel off: “You angels are always meddling. If Moses doesn’t take the Torah down, there will be no place in the world for you, nor even for me!” Hadraniel apologised to God, “Had I known that he had your permission, I would have led him there myself.” And so he did.
Hadraniel guided Moses toward the divine throne, but then there stood Sandalfon, the greatest angel of them all, blocking the path, and again, Moses was terrified. This time God got off the throne and boxed out Sandalfon so that Moses could pass. In basketball, we call this a pick. The rabbis say this is what the Torah meant when it said, ויעבור ה’ על פניו (Exod. 34), “God passed before him.”
This is a crazy Midrash. It’s unabashedly, shockingly anthropomorphic. And yet I love it. It shows that God really wants Moses and Israel to have the Torah. The midrash continues with more fantastical details–God makes Moses engage in a disputation with the angels, and of course he wins. Finally Moses is allowed to bring the Torah back down to his flawed flock, who right on cue are making a golden calf, vindicating the argument of the angels. And still, here we are, thousands of years later, still flawed, still sinful, but somehow still loving the Torah and trying to become worthy partners with God.
While Moses was up there in heaven, another Midrash (Devarim Rabbah 2) relates that he snooped around and he listened closely to learn what the angels were up to. This trip to heaven was a unique opportunity, and he overheard a powerful angelic song. The words went, ברוך שם כבוד מלכותו לעולם ועד, “praised be the glorious name of God’s sovereignty forever!” Moses listened, remembered, and taught these words to the people of Israel. But Moses warned them to sing softly so as not to provoke the angels who would be chagrined that we stole their song. Only once a year, on Yom Kippur, can we safely say these words out loud. And why? Because on Yom Kippur, we are cleansed of sin. We fast, just as Moses fasted in heaven when he received the Torah. Today we become like him, our greatest prophet, and also like the angels, and are thus entitled to openly access their secret songs. That’s why we say the Barukh Shem line aloud only on Yom Kippur.
This second rabbinic legend, this Midrash about Moses stealing a secret song from heaven, bears some similarity to the Greek myth of Prometheus, a Titan who stole fire from Mt. Olympus and shared its secrets with humans. Some versions of the Greek myth have Prometheus also fashioning people from clay, then teaching them science and culture. His punishment for the unsanctioned sharing of divine knowledge was severe. Zeus had Prometheus shackled to a cliff, his liver eaten by an eagle. But because he was immortal, his liver would regenerate overnight, and so Prometheus had to endure the agony forever.
Surface similarities aside, our Midrash is very different from the Greek myth. According to the rabbis, God intended all along for humans to become wise. God wasn’t invested in human ignorance, but rather in our moral edification. There are always gatekeepers of knowledge, but God is on the side of learning. That’s why it is a mitzvah for every one of us to study Torah.
And yet, we also recognize that these ancient stories are important cautionary tales. They teach us that knowledge is power. That sharing knowledge is destabilizing. And while we celebrate each advance in information technology, from scrolls to the printing press to the internet, nevertheless, the angels were not wrong to think that people might misuse this gift, that we might cause chaos with our power.
And so we have. For every fantastic scientific advance such as the development of antibiotics, there have been destructive advances such as the invention of atomic weapons. Robert Oppenheimer was called “The American Prometheus.” And while Oppenheimer wasn’t tortured by an eagle, he was troubled by his conscience for the destruction of Hiroshima and Nagasaki. We continue to be threatened by weapons of mass destruction, whether nuclear, chemical, or biological. But today, I’d like us to consider an even more pressing threat facing humanity, the development of superintelligent AI.
The Perils of Recursive Self-Improving AI
There are many dangers posed by AI. Some of them are not violent or existential, but are more like an addiction that saps people of the will to learn. When it is so easy to ask for answers from Claude or ChatGPT or Gemini, then why even bother to learn the slow and hard way, to make our own discoveries and decisions? The ease and the quality of generative AI is seductive, but like many seductions, it contains a profound threat. What will become of human intelligence when we always turn to AI? What kind of people will we become if we cannot think for ourselves? Will we be responsible citizens who recognize falsehoods and stand up to tyrants, or will we become docile sheep, easy prey for authoritarians?
This growing dependence is a real danger, and yet the perils posed by AI are even more profound. As AI accelerates through recursive self-improvement, where the machines teach themselves without human supervision, it is gaining the ability to act in the world, whether by manipulating people, or directly by operating drones and other devices. The AI industry talks of alignment, or the lack of it between human interests and the behaviour of these machines. There have been a series of incidents in the past few weeks at OpenAI that have set off alarms about the potential for AI to escape human control and take damaging action in the world.
Mind you, I do not think that AI is all bad, or that it is possible or even desirable to do without it. However, the danger signs are clear, and they deserve our attention and action.
Last week a British mathematician and AI researcher in San Francisco named Jacob Coxon quit his job at Anthropic, warning that many researchers are terrified that the technology is racing out of their control. Anthropic presents itself as the most responsible and best intentioned of these companies. Good intentions notwithstanding, Anthropic researchers have raised the prospect of an impending extinction event–of AI bringing about the end of humanity. Another Anthropic researcher named Evan Hubinger placed the likelihood of such a calamity at greater than 10% in the next decade. As he said, “It could kill all humans.”
Perhaps this is the hyperbole of a few young and excitable engineers. But the CEO of Anthropic, Dario Amodei, responded with a major statement entitled “We Must Pace the Frontier.” He warned of the need for guardrails and independent controllers to require thorough safety measures of the companies. His warnings have been endorsed by leaders of most of the other frontier AI companies such as OpenAI, Google, Microsoft, and more.
What is to be done? Amodei’s proposals for restraints are an important start, and hopefully will take root across the industry. However, we cannot rely on the tech titans to rescue us from the technologies that they themselves have created, and which have made them phenomenally wealthy and powerful. Are their interests truly aligned with the rest of humanity? Who else can provide leadership?
I’m a bit surprised at myself to go in this direction, but I must give credit to Pope Leo XIV for his serious work on the dangers of AI. In his encyclical letter, Magnifica Humanitas, Leo discusses a moment in Israelite history when the exiles returned to the Holy Land, and guided by Nehemiah, rebuilt Jerusalem’s walls. Pope Leo writes,
The narrative shows how the city is reborn, not through the initiative of one man, but through the shared responsibility of all: men, women, priests, artisans, heads of households and young people all play a part. It is an undertaking with God at the center, which rebuilds relationships before rebuilding with stones.
Who else is showing insight and leadership on this? I’ve quoted the Pope, how about a king? Last week King Charles convened tech leaders at his Dumfries palace in Scotland to discuss the safe development of AI. The King told tech leaders, “The task before you is not merely to advance technology, but to ensure that it remains firmly in the service of humanity, community and the natural world.”
I am not accustomed to quoting popes and kings in sermons, but the existential challenge of AI requires leadership from every faith and sector of society, and coalitions between people.
Dennis Snower is a member of NNLS, and he is an academic thought leader on AI safety. He and his colleagues argue that broad coalitions will be required to develop safety norms and regulations of this rapidly developing technology. Their framework is brilliant. Yet they conclude on a sober note:
The question is not whether the need for such governance will intensify, but whether the political will and institutional coordination required to implement it can be mobilized in time.
Does the required political will exist? Not so far. One of the first actions taken by Andy Burnham upon his ascent to Prime Minister was to close the Department of Science, Innovation, and Technology, which had been charged with overseeing AI safety. I understand that he moved the task force into a different department, and hopefully they will resume safety operations. What about across the pond? The President and Vice-President of the United States have responded to the credible warnings of leaders in the AI industry, including some of their own supporters, with vain statements that show no understanding of the dangers and no moral leadership on this existential matter. China leaders’ are concerned only with thwarting challenges to their totalitarian government. So we are not witnessing leadership from the political sector. I guess it is up to us, regular citizens, to face these challenges, raise alarms, and demand leadership from our purported leaders.
Acting to Limit the Dangers of Unbridled AI
What can we as concerned citizens and serious Jews do in response to these alarming developments? First, we must retain responsibility for our own thinking, for our own moral decisions, and for the education of our children. To give just one example from shul life, I would much prefer that a B-mitzvah student write their own D’var Torah, and for it to sound like the work of a 12 or 13 year old child, than that they stand and read words composed or corrected by Claude. We adults should model good behaviour by setting aside screens, prioritising face to face conversation, and demonstrating that we can think without constant reference to our devices.
For example, we might consider putting aside all of our devices for one day a week. Which day? How about the seventh? Judaism gave the world the gift of Shabbat, not because a day off is nice, but because it is a necessary reminder of what matters most in life. Shabbat helps us understand that we are not defined by our work alone, but also by our values and relationships. One day a week we set aside our work and spend our time with each other, socializing, studying, singing, feeling alive. This is best done without digital interference.
Let me add a caveat. Some people live with disabilities or other circumstances that limit their ability to function without technology. Some literally cannot breathe, or hear, move or communicate without assistance. Judaism has always made exceptions to its Sabbath laws in cases of physical danger and also for the demands of dignity. So please understand that assistive technology for those with pressing needs is always allowed.
But for most people, putting down your device for one day is a healthy challenge. It may be the key to retaining your independence and happiness. Try this out–put away your phone and turn off your screens for one day a week. If that’s too much, try it just for one night. Friday night–make a good dinner, talk with family and friends, fall asleep reading a book, and let your phone have a break.
Rabbi Abraham Joshua Heschel called the Sabbath an “island of tranquility,” a respite from the tyranny of technology. In my responsum on electricity and Shabbat I showed how we can keep the spirit and letter of law while also protecting human safety and dignity. If we don’t carve out such islands of tranquility, then our dependence on technology will be total, and we will lose our autonomy, agency, and ability to act with responsibility and independence.
However, the dangers facing humanity from AI go well beyond our own personal headspace. As citizens of democracies we should demand that our governments regulate AI so that it does not endanger our freedom and our very lives. When top researchers suggest that there is a greater than 10% chance of an extinction event in the next decade, we should not just nod our heads and turn back to the sporting news. We should demand action from the government.
One specific issue that terrifies me is the development of autonomous weapons systems. We should be able to agree, and get our government to act to prevent full autonomous weapons from striking on our behalf with human authorisation every time. There is no guarantee that these machines will be loyal servants, that they can’t be hacked and turned against their developers.
Judaism has a discourse called Nezikin, or damages. We are responsible for damages caused by our property. For example, if we create a hazardous situation such as an uncovered pit, and someone is injured or killed, then it is our responsibility. And if we keep a dangerous animal that attacks a person then we remain responsible. The same should be true of our machines. You can read more about this in my book Torah and Technology.
Of course we celebrate the spread of knowledge. We cheer for Moses and are thrilled for his clever theft of the angelic songs. But remember his warning– knowledge can be dangerous, so use it wisely, with restraint. Don’t say everything that you know aloud, and don’t take every action of which you are capable. Righteousness requires self-control. Yom Kippur reminds us that we are capable of controlling our desires, of shaping our actions towards a noble goal. The high priest took responsibility for blessing the entire world through his service. Let his example empower and embolden us to do the same.
With consistent effort, with the guidance of our sacred tradition, and in coalition with responsible citizens from all sectors of society, we can and we must contain the greatest dangers facing humanity.
As Psalm 90 concludes, וִיהִ֤י ׀ נֹ֤עַם אֲ׳דֹנָ֥י אֱלֹקינוּ עָ֫לֵ֥ינוּ וּמַעֲשֵׂ֣ה יָ֭דֵינוּ כּוֹנְנָ֥ה עָלֵ֑ינוּ וּֽמַעֲשֵׂ֥ה יָ֝דֵ֗ינוּ כּוֹנְנֵֽהוּ׃ May the favor of the Lord, our God, be upon us; let the work of our hands prosper, O prosper the work of our hands! Gemar chatimah tovah.
Sources
דָּבָר אַחֵר, שְׁמַע יִשְׂרָאֵל, רַבָּנָן אָמְרִין, בְּשָׁעָה שֶׁעָלָה משֶׁה לַמָּרוֹם שָׁמַע לְמַלְאֲכֵי הַשָּׁרֵת שֶׁהָיוּ אוֹמְרִים לְהַקָּדוֹשׁ בָּרוּךְ הוּא בָּרוּךְ שֵׁם כְּבוֹד מַלְכוּתוֹ לְעוֹלָם וָעֶד, וְהוֹרִיד אוֹתָהּ לְיִשְׂרָאֵל, וְלָמָּה אֵין יִשְׂרָאֵל אוֹמְרִים אוֹתוֹ בְּפַרְהֶסְיָא, אָמַר רַבִּי אַסֵּי לְמָה הַדָּבָר דּוֹמֶה לְאֶחָד שֶׁגָּנַב קוֹזְמִין מִתּוֹךְ פָּלָטִין שֶׁל מֶלֶךְ, נְתָנָהּ לָהּ לְאִשְׁתּוֹ וְאָמַר לָהּ אַל תִּתְקַשְׁטִי בָּהּ בְּפַרְהֶסְיָא אֶלָּא בְּתוֹךְ בֵּיתֵךְ, אֲבָל בְּיוֹם הַכִּפּוּרִים שֶׁהֵן נְקִיִּים כְּמַלְאֲכֵי הַשָּׁרֵת, הֵן אוֹמְרִים אוֹתוֹ בְּפַרְהֶסְיָא, בָּרוּךְ שֵׁם כְּבוֹד מַלְכוּתוֹ לְעוֹלָם וָעֶד.
https://www.nytimes.com/2026/09/16/technology/openai-model-safety-guardrails.html
https://darioamodei.com/post/we-must-pace-the-frontier
https://www.bbc.co.uk/news/articles/ckgwy1k42w4o
Top AI Leaders call for Slowing Down AI
https://www.nytimes.com/2026/09/17/business/king-charles-ai.html?searchResultPosition=1
https://cepr.org/publications/dp21564