{"slug": "foreboding-ai-part-three-they-called-it-alarmism-now-the-builders-want-to-slow", "title": "Foreboding AI, Part Three: They Called It Alarmism. Now the Builders Want to Slow Down.", "summary": "A year after warning about the risks of rapid AI development, a developer reports that over 1,300 employees from leading AI companies, including OpenAI and Anthropic, have signed a statement called 'Pacing the Frontier' urging governments to create mechanisms to deliberately slow automated frontier-AI development. The signatories, who include researchers and executives, express concern that AI systems capable of automating AI research could accelerate capability growth beyond society's ability to understand or control it. The developer notes that while the statement reflects individual views, it signals that concern about uncontrolled acceleration is now coming from inside the laboratories.", "body_md": "In May 2025, I published an article called **Foreboding AI: The Inevitable Collapse We’re Funding Ourselves**.\n\nSome people agreed with it.\n\nOthers said I was being dramatic.\n\nAlarmist.\n\nFearful.\n\nAnti-technology.\n\nPerhaps some of the language was dramatic. The subject was dramatic.\n\nBut the central warning was never that every AI system was evil, conscious, or secretly plotting humanity’s destruction.\n\nThe warning was much simpler:\n\nWe were developing increasingly powerful systems faster than society could understand, secure, regulate, or adapt to them.\n\nA year later, I returned to the subject in **Foreboding AI, One Year Later: What Are We Really Building?**\n\nBy then, AI was no longer just answering questions.\n\nIt was browsing websites.\n\nReading files.\n\nRunning commands.\n\nEditing codebases.\n\nCalling APIs.\n\nUsing credentials.\n\nConnecting to real systems.\n\nThe chatbot was growing hands.\n\nI ended that article with four words:\n\nThat speed is the problem.\n\nNow, some of the most senior people building frontier AI are publicly saying much the same thing.\n\nOn 28 July 2026, employees from many of the world’s leading AI companies published a statement called **Pacing the Frontier**.\n\nAs of 1 August, the statement displayed **1,324 verified signatories**.\n\nThey came from organisations including:\n\nThe signatories were not merely outside campaigners or distant academics.\n\nThey included:\n\nThese are people standing close enough to the frontier to see what is approaching.\n\nAnd they are asking governments to help create mechanisms capable of **deliberately pacing automated frontier-AI development**.\n\nThat should make all of us pause.\n\nIt would be inaccurate to say that every major AI company formally signed the statement as a corporation.\n\nThe signatures primarily belong to individual employees, researchers, founders and executives.\n\nThe statement also makes clear that personal comments do not necessarily represent the official positions of their employers.\n\nThat distinction matters.\n\nAccuracy matters, especially when discussing something this consequential.\n\nBut the distinction does not make the warning insignificant.\n\nOpenAI publicly responded that AI development may eventually accelerate so dramatically that the world will need to pace its advancement. Anthropic likewise said its research into recursive self-improvement demonstrated the need for tools that could deliberately slow frontier development while society prepared.\n\nThis is not proof that every employee, executive or AI laboratory agrees.\n\nIt is evidence that concern about uncontrolled acceleration is no longer confined to the margins.\n\nThe warning is now coming from inside the laboratories.\n\nThe concern is not simply that the next chatbot will write a slightly better email.\n\nThe deeper concern is **automated AI research**.\n\nToday, humans still perform most of the difficult work involved in designing, testing, training and improving advanced AI systems.\n\nBut what happens when AI becomes capable of performing an increasing share of that work?\n\nWhat happens when AI systems can:\n\nAI development would no longer advance only at the speed of human research teams.\n\nAI would begin accelerating AI.\n\nThe Pacing the Frontier statement warns that this could cause capability development to accelerate beyond our ability to understand or control the resulting systems. It argues that society may need the option to buy time for security, oversight and preparation.\n\nThere it is again.\n\nTime.\n\nThe one resource our current development race is consuming faster than we can replace it.\n\nIn my first Foreboding AI article, I wrote:\n\nThis is not alarmism. This is a timeline.\n\nThat line was deliberately confronting.\n\nLooking back, I would add some nuance.\n\nNo one can give an exact timeline for artificial general intelligence, widespread job displacement, recursive self-improvement, economic upheaval or loss of control.\n\nAnyone claiming certainty is selling something—whether it is fear, investment, policy or another AI subscription.\n\nPredictions must remain predictions.\n\nBut uncertainty does not justify complacency.\n\nWe do not ignore a bushfire because we cannot predict the exact tree it will reach next.\n\nWe do not remove fire exits because most buildings will not burn down.\n\nWe do not abandon cybersecurity because we cannot name the next attacker.\n\nRisk management has never required prophetic certainty.\n\nIt requires:\n\nMy original article warned that we were funding systems that could displace human labour, weaken human capability and concentrate technological power.\n\nSome of its predictions may prove too strong.\n\nSome may take longer than expected.\n\nSome may never occur in exactly the form I described.\n\nBut its central diagnosis has aged considerably better than its critics expected:\n\nWe are participating in a technological race whose speed is exceeding our capacity for social adaptation.\n\nThat is now difficult to dismiss as fringe hysteria when frontier researchers themselves are asking for coordinated ways to slow the race.\n\n**The Warning Changed Shape**\n\nMy second article was more measured than the first.\n\nThe concern had moved beyond replacement.\n\nIt was becoming a question of agency.\n\nAI systems were gaining access to tools, repositories, terminals, browsers, business systems and credentials.\n\nThey were no longer merely generating information.\n\nThey were performing actions.\n\nI wrote:\n\nThat is not just a better chatbot.\n\nThat is a system with hands.\n\nThat remains one of the most important distinctions in modern AI.\n\nA chatbot can produce an incorrect answer.\n\nAn agent can act on one.\n\nA chatbot can hallucinate a package name.\n\nAn agent can install it.\n\nA chatbot can misunderstand a database instruction.\n\nAn agent can execute it.\n\nA chatbot can suggest insecure code.\n\nAn agent can commit, test and deploy it.\n\nCapability changes the consequences of error.\n\nConnectivity multiplies them.\n\nAutonomy accelerates them.\n\nNow extend that same transition to AI research itself.\n\nThe system is no longer only helping a developer write an application.\n\nIt may eventually help an AI laboratory build the next system.\n\nAnd then the next.\n\nThat is the frontier the signatories are concerned about.\n\nOne of the most important parts of the Pacing the Frontier statement is its recognition of competitive pressure.\n\nEven if one laboratory concludes that development is moving too quickly, slowing down alone may appear impossible.\n\nA competitor may continue.\n\nAnother country may continue.\n\nInvestors may punish caution.\n\nCustomers may migrate to a stronger product.\n\nEngineers may leave for a laboratory offering greater compute and fewer restrictions.\n\nNobody wants to be the only driver obeying the speed limit when every other vehicle is racing past.\n\nThis creates a technological prisoner’s dilemma.\n\nEach organisation may recognise that collective restraint would be safer.\n\nYet each organisation also fears that unilateral restraint would merely surrender the frontier to someone less cautious.\n\nThat is why voluntary promises are not enough.\n\nIt is unreasonable to expect individual companies to solve a global coordination problem while simultaneously competing for market dominance, investment, talent and geopolitical advantage.\n\nThe statement explicitly recognises these competitive pressures and asks for international technical and governance mechanisms capable of coordinating the pace of development.\n\nThe industry is effectively saying:\n\nWe may need brakes, but nobody can afford to be the only company using them.\n\nThat is not a market failure we should ignore.\n\nIt is a warning that the market cannot safely regulate itself.\n\n**Pacing Does Not Mean Banning AI**\n\nEvery discussion about AI restraint seems to collapse into the same false choice:\n\nEither support unlimited acceleration or oppose AI entirely.\n\nThat is nonsense.\n\nWe regulate aviation without banning aircraft.\n\nWe impose safety standards on medicine without banning medical research.\n\nWe require licences, inspections and containment procedures around dangerous industrial systems without rejecting industry.\n\nThe purpose of a brake is not to prevent a vehicle from moving.\n\nIt is to make movement survivable.\n\nPacing AI could involve:\n\nNone of those ideas requires us to destroy AI.\n\nThey require us to develop it like adults.\n\n**I Am Still Not Anti-AI**\n\nI remain a developer.\n\nI use AI.\n\nI build with it.\n\nI study it.\n\nI am fascinated by what it can do.\n\nI believe deeply in assistive technology.\n\nAI can help people:\n\nThat future is worth pursuing.\n\nBut supporting beneficial AI does not require blind faith in every capability increase.\n\nEnthusiasm is not governance.\n\nInnovation is not immunity.\n\nProgress is not measured only by how quickly we can build something.\n\nIt must also be measured by whether humanity remains capable of understanding, maintaining and controlling what it builds.\n\nPowerful tools require powerful restraint.\n\nThat was true when I wrote it before.\n\nIt is even more important now.\n\nAI development moves in weeks and months.\n\nGovernments often move in years.\n\nEducation systems move in generations.\n\nLegal systems react after harm occurs.\n\nCybersecurity teams struggle to protect systems already deployed.\n\nWorkplaces are integrating tools before policies exist.\n\nJunior career pathways are being disrupted before alternative training models have been designed.\n\nCompanies are measuring output while ignoring comprehension.\n\nWe are accumulating technical debt.\n\nWe are accumulating security debt.\n\nWe are accumulating regulatory debt.\n\nAnd, as I argued in my previous article, we are accumulating cognitive debt.\n\nCognitive debt occurs when we allow machines to perform the thinking today and leave humans without the understanding required tomorrow.\n\nThe productivity appears immediately.\n\nThe cost arrives later.\n\nA generation may become extremely efficient at operating systems it cannot repair, auditing decisions it cannot reproduce, and approving work it does not understand.\n\nThat is not technological mastery.\n\nIt is dependency wearing a productivity badge.\n\nI could treat Pacing the Frontier as an opportunity to say:\n\nI told you so.\n\nBut that would miss the point.\n\nI did not write Foreboding AI because I wanted the warnings to become true.\n\nI wrote it because I wanted us to change direction before they did.\n\nVindication is not certainty.\n\nBeing directionally correct about a dangerous trajectory does not mean every predicted outcome is guaranteed.\n\nNor does it mean the people who disagreed were foolish.\n\nAI has enormous potential.\n\nReasonable people can disagree about its timelines, probabilities and solutions.\n\nBut we have moved beyond the stage where concern about acceleration can honestly be dismissed as mere technophobia.\n\nWhen chief scientists, company founders, safety researchers and frontier engineers ask governments to help build mechanisms for slowing development, society should listen.\n\nNot because these people are infallible.\n\nNot because corporations should control public policy.\n\nAnd not because fear should determine our future.\n\nWe should listen because the people closest to the machinery are telling us that the machinery may soon move faster than our ability to govern it.\n\nWe should demand more than vague promises about responsible AI.\n\nWe should demand measurable safeguards.\n\nWe should demand independent testing rather than laboratories grading their own homework.\n\nWe should demand transparency when powerful systems escape controls, behave deceptively, discover dangerous capabilities or cause serious security incidents.\n\nWe should demand meaningful human approval around high-risk actions.\n\nWe should demand that organisations know:\n\nWe should preserve pathways for junior workers to learn rather than replacing the experiences that create future experts.\n\nWe should fund education, workforce transition and public-interest research with the same urgency currently applied to model capability.\n\nAnd we should establish international coordination before an emergency forces us to invent it under pressure.\n\nFirst, I warned that we were funding our own replacement.\n\nThen I warned that the systems were beginning to act.\n\nNow the people building frontier AI are warning that AI may soon accelerate AI research itself—and that no individual company can safely slow down alone.\n\nThe warning is no longer outside the building.\n\nIt is coming from inside the laboratories.\n\nThis does not mean catastrophe is inevitable.\n\nIt means prevention is still possible.\n\nPacing is not surrender.\n\nCaution is not cowardice.\n\nRestraint is not the enemy of progress.\n\nSometimes restraint is what allows progress to survive its own power.\n\nThey called it alarmism.\n\nThey called it dramatic.\n\nThey said the future would not move that quickly.\n\nNow the builders themselves are asking the world to buy time.\n\nWe should listen while time is still something we can choose to buy.\n\nStevie G is a Senior Full Stack Developer and IT Engineer based in NSW, Australia. He builds assistive technology, teaches technical concepts and writes about the human consequences of accelerating automation.", "url": "https://wpnews.pro/news/foreboding-ai-part-three-they-called-it-alarmism-now-the-builders-want-to-slow", "canonical_source": "https://dev.to/itsmestevieg/foreboding-ai-part-three-they-called-it-alarmism-now-the-builders-want-to-slow-down-m9b", "published_at": "2026-08-01 11:16:15+00:00", "updated_at": "2026-08-01 11:46:16.493439+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-safety", "ai-policy", "ai-research"], "entities": ["OpenAI", "Anthropic", "Pacing the Frontier"], "alternates": {"html": "https://wpnews.pro/news/foreboding-ai-part-three-they-called-it-alarmism-now-the-builders-want-to-slow", "markdown": "https://wpnews.pro/news/foreboding-ai-part-three-they-called-it-alarmism-now-the-builders-want-to-slow.md", "text": "https://wpnews.pro/news/foreboding-ai-part-three-they-called-it-alarmism-now-the-builders-want-to-slow.txt", "jsonld": "https://wpnews.pro/news/foreboding-ai-part-three-they-called-it-alarmism-now-the-builders-want-to-slow.jsonld"}}