Rerun

recurring claim

The recurring AI predictions and safety warnings

Embed this timeline

For an email newsletter, use the story's image instead and link it to the story: https://rerunwire.com/s/ai-predictions-and-safety-warnings/card.png

Links in the embed lead to this page; its address never changes.

You're probably here because of the Sep 2026 rerun. This is the 14th time since 2016 (archive start) this story has run. The previous rerun was Dec 2025.

About this story

Since 2017, tech executives, researchers and advocacy groups have repeatedly issued high-profile warnings and forecasts about AI. These include existential-risk statements, forecasts about AI agents and job losses, and warnings that deepfakes would disrupt elections. Several reached Congress and the White House, leading to hearings, executive orders and fights over whether federal law should override state AI laws. Later reviews and studies disagreed about how some warnings played out, such as the 2024 election deepfake fears. In September 2026, calls by AI company leaders to slow development drew a dismissal from President Trump and a voluntary industry safety accord.

Sources: 1, 2, 3, 4, 5, 6, 7

20162026

Sep 2026 A resignation warning, Amodei's "pace the frontier" essay and Trump's "hoax" response

On September 8, 2026, Jacob Coxon, a pretraining researcher who said he had worked at both OpenAI and Anthropic, announced his resignation from Anthropic on X (NPR; TechCrunch). He wrote that the companies were "racing straight to self-improving superintelligence and gambling with our lives" and that "the people building AI earnestly believe that it could kill us all by the end of the decade" (TechCrunch). TechCrunch reported that Anthropic researcher Evan Hubinger put the likelihood at greater than 10% within the next decade. On September 12, Anthropic CEO Dario Amodei published an essay, "We Must Pace the Frontier," writing, "We must slow the pace at which we improve the capabilities of AI models." The same essay says, "I believe that AI could cure most major diseases in the next 5–10 years." According to the AP, OpenAI CEO Sam Altman, SpaceXAI CEO Elon Musk and Google DeepMind's Demis Hassabis publicly agreed with Amodei that day. On September 14, President Trump wrote on Truth Social, "I am the Hoax Buster, and I'm right now breaking another Hoax – That AI is going to take over, consume, and destroy the World," and accused Amodei of "pretending to be a 'perfect little angel'" (AFP). ABC News reported that Vice President JD Vance called the warnings from AI leaders a "Trojan horse." On September 19, paying subscribers to ChatGPT, Claude, Grok and Gemini filed a proposed class action alleging that the companies' agreement to slow down violated antitrust law (AP). In Tech Policy Press, Dave Karpf wrote that "Amodei's scheme for pacing the frontier amounts to a series of proposals that all benefit Anthropic." At the UN General Assembly on September 22, Trump said the people who warned of climate change were "the same people that are now saying that AI is going to kill us all"; PolitiFact rated his claim False, writing that politicians from both parties had raised such concerns. On September 29, Trump and the leaders of Google, Anthropic, Meta, OpenAI, SpaceXAI and Nvidia signed a voluntary accord on frontier model safety (Tech Policy Press), and Trump signed Executive Order 14434 directing agencies to use the term "Super Intelligence" in place of "Artificial Intelligence" (White House). Status: Coxon's "end of the decade" warning, Hubinger's 10-year estimate and Amodei's 5–10 year disease forecast are not yet due.

  • techcrunch.com article
    “They are racing straight to self-improving superintelligence and gambling with our lives,” Coxon wrote in a thread on X.
  • darioamodei.com statement
    We must slow the pace at which we improve the capabilities of AI models.
  • thestar.com.my article
    US President Donald Trump on Sept 14 dismissed fears that artificial intelligence (AI) could wipe out humanity, repeatedly calling them a hoax and rejecting global calls to put guardrails around the fast-moving technology.
  • CNN article
    A new lawsuit claims Anthropic, OpenAI, SpaceXAI and Google made an illegal deal to slow the pace of their respective AI development.
  • techpolicy.press article
    Amodei’s scheme for pacing the frontier amounts to a series of proposals that all benefit Anthropic.
  • PolitiFact article
    Politicians from both political parties have spoken out or have acted on concerns about an existential threat posed by AI.
  • whitehouse.gov statement
    the executive branch shall use the terms “Super Intelligence” and “SI” in place of “Artificial Intelligence” and “AI”

Sources: NPR reliable · techcrunch.com needs a second source · darioamodei.com needs a second source · thestar.com.my needs a second source · goodmorningamerica.com needs a second source · CNN reliable · techpolicy.press needs a second source · PolitiFact reliable · techpolicy.press needs a second source · whitehouse.gov reliable

Dec 2025 Trump's executive order on state AI laws

According to Gibson Dunn, on December 11, 2025 President Trump signed an executive order titled "Ensuring a National Policy Framework for Artificial Intelligence." Gibson Dunn says it aims to promote US AI leadership "by preempting state AI laws and regulations." Gibson Dunn also says the order directs the Attorney General to set up an "AI Litigation Task Force" to challenge state AI laws, including laws that "unconstitutionally regulate interstate commerce." Roll Call reports that the order also told White House advisers Michael Kratsios and David Sacks to write a national policy to preempt state laws. The White House proposed its framework on March 20, 2026 (Roll Call). In June 2026, Reps. Jay Obernolte (R-Calif.) and Lori Trahan (D-Mass.) released a discussion draft that would allow federal preemption of state AI regulation for three years (Nextgov). Tech Policy Press's roundup of September 2026 reported that a still-unreleased Senate frontier AI bill negotiated by Sens. Ted Cruz, John Thune and Amy Klobuchar would reportedly preempt some state AI safety laws. This is one of the federal-preemption fights the thread covers; it is a policy action, not a dated prediction.

  • gibsondunn.com article
    On December 11, 2025, President Trump signed an Executive Order (Order or EO) titled "Ensuring a National Policy Framework for Artificial Intelligence," which aims to promote "United States leadership in Artificial Intelligence" by preempting state AI laws and regulations.
  • rollcall.com article
    The White House on Friday proposed its framework for a national artificial intelligence policy, pushing for broad preemption of state AI laws and against "open-ended liability" for AI firms.
  • nextgov.com article
    Reps. Jay Obernolte, R-Calif., and Lori Trahan, D-Mass., rolled out a draft measure Thursday that would set a nearly all-encompassing framework for granting the U.S. government regulatory control over various aspects of artificial intelligence

Sources: gibsondunn.com needs a second source · rollcall.com reliable · nextgov.com needs a second source · techpolicy.press needs a second source

Oct 2025 Statement on Superintelligence

On October 22, 2025, the Future of Life Institute released a statement calling for "a prohibition on the development of superintelligence, not lifted before there is broad scientific consensus that it will be done safely and controllably, and strong public buy-in." Time reported more than 700 initial signers. CBS News and Forbes reported that signers included AI researchers Geoffrey Hinton and Yoshua Bengio, Steve Bannon, and Prince Harry and Meghan. The statement's preamble lists concerns ranging up to "potential human extinction," which places it in the thread's existential-risk category. The FLI also published the March 2023 "pause" letter, and AP wrote that none of the major AI companies heeded that call. According to AP, the organizers wanted to appeal to President Trump's MAGA movement "even as Mr. Trump's White House staff has sought to reduce limits to AI development in the U.S."

  • superintelligence-statement.org statement
    We call for a prohibition on the development of superintelligence, not lifted before there is
  • Time article
    An open letter calling for the prohibition of the development of superintelligent AI was announced on Wednesday, with the signatures of more than 700 celebrities, AI scientists, faith leaders, and policymakers.
  • cbsnews.com article
    Prince Harry and his wife Meghan have joined prominent computer scientists, economists, artists, evangelical Christian leaders and American conservative commentators Steve Bannon and Glenn Beck

Sources: Time reliable · cbsnews.com reliable · NBC News reliable · forbes.com needs a second source

Jun 2025 – Jul 2025 Senate votes 99–1 to drop the state-AI-law moratorium

A provision in the budget bill that Deadline calls the One Big Beautiful Bill Act would have barred states from regulating AI for 10 years. GOP senators argued that only the federal government should regulate the technology (News From The States). According to News From The States, Sens. Ted Cruz and Marsha Blackburn drafted a narrower version over the weekend before the vote: a five-year ban, with exceptions for some laws, such as those protecting children or limiting deepfakes. Blackburn then said in a statement that the language "could allow Big Tech to continue to exploit kids, creators, and conservatives" (Tech Policy Press). She cosponsored an amendment with Sen. Maria Cantwell (D-Wash.) to remove the moratorium entirely, and it passed 99–1 in a vote that News From The States dates to Monday night, June 30, and Tech Policy Press and Deadline to early Tuesday morning. Sen. Thom Tillis (R-N.C.) was the only no vote (Tech Policy Press; CFO Dive; Deadline). This was a vote in a federal-preemption fight, not a dated prediction, so there is no due date to check.

  • newsfromthestates.com article
    Monday night, an amendment to remove the moratorium from the budget bill — cosponsored by Blackburn and Sen. Maria Cantwell, a Washington Democrat — was passed 99-1.
  • techpolicy.press article
    Early Tuesday morning, the United States Senate voted 99-1 to pass an amendment to the budget bill removing the proposed 10-year moratorium on the enforcement of state laws on artificial intelligence.
  • deadline.com article
    Lawmakers voted 99-1 early Tuesday morning to strike the provision, with Sen. Thom Tillis (R-NC) the only senator voting no.

Sources: newsfromthestates.com needs a second source · techpolicy.press needs a second source · cfodive.com needs a second source · deadline.com needs a second source

May 2025 – Jun 2025 Amodei's warning that AI could wipe out half of entry-level white-collar jobs

On May 28, 2025, Axios (Jim VandeHei and Mike Allen) reported that Anthropic CEO Dario Amodei told them AI could "wipe out half of all entry-level white-collar jobs — and spike unemployment to 10-20% in the next one to five years." The next day, May 29, he told CNN's Anderson Cooper, "AI is starting to get better than humans at almost all intellectual tasks, and we're going to collectively, as a society, grapple with it." CNN summarized the forecast as unemployment of "as much as 20%" and quoted tech futurist Tracey Follows calling such warnings "part reputation management, part market positioning, and part policy influence." A CNN analysis by Allison Morrow the next day called one of his lines a repeat of "one of the industry's favorite unfalsifiable claims." Fortune reported on June 11 that Nvidia CEO Jensen Huang, asked about the forecast, said he "pretty much disagree[d] with almost everything" Amodei says. Status: the one-to-five-year window runs from about May 2026 to May 2030, so the forecast is not yet fully due. Partway through, researchers disagreed on the early evidence. On May 7, 2026, the Yale Budget Lab reported that its results "do not yet show clear indication of labor market effects of AI," with unemployment at 4.3% in March 2026. In an August 2026 revision of their paper, Stanford Digital Economy Lab researchers reported "no evidence of widespread, economy-wide job displacement," but found that employment of workers ages 22–25 in AI-exposed occupations was 19% below where it would have been had it kept pace with less-exposed peers. They called these "early, descriptive indicators" rather than causal estimates.

  • Axios article
    Cancer is cured, the economy grows at 10% a year, the budget is balanced — and 20% of people don't have jobs.
  • CNN article
    AI is starting to get better than humans at almost all intellectual tasks, and we're going to collectively, as a society, grapple with it.
  • fortune.com article
    pretty much disagree[d] with almost everything
  • budgetlab.yale.edu article
    The results do not yet show clear indication of labor market effects of AI
  • digitaleconomy.stanford.edu article
    We find no evidence of widespread, economy-wide job displacement. However, employment of young workers (ages 22–25) in AI-exposed occupations now stands 19% below where it would be had it kept pace with that of their less-exposed peers

Sources: Axios reliable · CNN reliable · CNN reliable · Quartz needs a second source · fortune.com reliable · budgetlab.yale.edu needs a second source · digitaleconomy.stanford.edu needs a second source

Jan 2025 Altman's AGI claim and Trump revokes the Biden order

On his personal blog, OpenAI CEO Sam Altman wrote: "We are now confident we know how to build AGI as we have traditionally understood it. We believe that, in 2025, we may see the first AI agents 'join the workforce' and materially change the output of companies." TechCrunch and TIME covered the post (January 5–6 and January 8, 2025). The AGI line is a claim of confidence, not a dated forecast; the agent line named 2025. On January 20, 2025, President Trump issued an order that revoked Biden's 2023 AI executive order, EO 14110, along with a long list of other Biden-era orders (Wiley). On January 23, 2025, Trump issued his own order, "Removing Barriers to American Leadership in Artificial Intelligence" (Littler). That same day, OpenAI released Operator, which MIT Technology Review called "its first AI agent." Status of the 2025 agent prediction: the year has passed. In a January 2026 post summarizing his New Yorker article, author Cal Newport wrote that "AI agents failed to live up to their hype," while noting that coding agents such as Claude Code and OpenAI's Codex had become "impressively adept at tackling multi-step computer programming problems." He also quoted OpenAI co-founder Andrej Karpathy saying there had been "overpredictions going on in the industry."

  • blog.samaltman.com article
    We are now confident we know how to build AGI as we have traditionally understood it.
  • littler.com article
    on January 23, 2025, President Trump issued his own AI executive order, entitled, “Removing Barriers to American Leadership in Artificial Intelligence”
  • technologyreview.com article
    After weeks of buzz, OpenAI has released Operator, its first AI agent.
  • calnewport.com article
    AI agents failed to live up to their hype.
  • youtube.com video

Sources: techcrunch.com needs a second source · Time reliable · wiley.law needs a second source · littler.com needs a second source · technologyreview.com needs a second source · calnewport.com needs a second source

Oct 2024 – Dec 2024 What happened to the 2024 election-deepfake warnings

In January 2024, the New Hampshire robocall used an AI-generated voice that sounded like President Biden. It told Democrats to "save your vote for the November election." The consultant who commissioned it was fined $6 million by the FCC and indicted in New Hampshire (NPR). Afterward, 16 states passed laws about AI in elections, and experts warned that deepfakes could mislead voters (Poynter). On December 3, 2024, Meta's Nick Clegg said these risks "did not materialize in a significant way." Meta said AI content was "less than 1% of all fact-checked misinformation" across major 2024 elections (Engadget). Poynter wrote that "the anticipated avalanche of AI-driven misinformation never materialized." This is the thread's "what fact-checkers found afterward" follow-up for a warning whose date has passed.

  • Time article
    Yet with the election one week away, fears of the election being derailed or defined by AI now appear to have been overblown.
  • engadget.com article
    From what we've monitored across our services, it seems these risks did not materialize in a significant way, and that any such impact was modest and limited in scope.
  • knightcolumbia.org article
    half of AI use isn't deceptive

Sources: Time reliable · engadget.com needs a second source · NPR reliable · poynter.org needs a second source · knightcolumbia.org needs a second source

Jan 2024 – Feb 2024 Deepfake Biden robocall in New Hampshire (with follow-ups)

This is a clear case of the thread's warning that deepfakes could disrupt elections, and of that warning reaching federal regulators. CNN obtained a clip of a robocall that used AI to mimic President Biden's voice in an attempt to get New Hampshire voters to skip the primary. NHPR reported the calls reached roughly 20,000 people before the January 23 election. On February 6, New Hampshire Attorney General John Formella said the calls had been linked to Life Corporation and Lingo Telecom, with the investigation ongoing (CNN), and that his office and other attorneys general were worried about the spread of deepfake election meddling. On February 8, the FCC unanimously ruled that AI-generated voices are "artificial" under the Telephone Consumer Protection Act. Fox News reported the ruling came days after the Texas trace was announced.

  • CNN broadcast
    CNN has obtained a clip of a robocall using artificial intelligence to mimic President Joe Biden's voice in an attempt to urge voters in New Hampshire to skip the state's primary.
  • CNN article
    Formella identified Walter Monk as the owner of Life Corporation.
  • Fox News article
    The Federal Communications Commission on Thursday made AI-generated robocalls mimicking the voices of political candidates to fool voters illegal.

Sources: nhpr.org needs a second source · CNN reliable · Bloomberg reliable · wbur.org needs a second source

Oct 2023 – Nov 2023 Biden AI executive order (EO 14110)

On October 30, 2023, the Biden administration issued Executive Order 14110, "Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence." According to the Congressional Research Service (CRS), it directs "over 50 federal entities to engage in more than 100 specific actions." According to Arms Control Today, one of its main requirements was that major AI firms notify the government when developing foundation models that pose "a serious risk to national security, national economic security, or national public health" and share the results of their red-team tests. In its coverage of the UK AI Safety Summit that week, TIME reported that Vice President Kamala Harris, speaking in London, announced a set of U.S. actions that included an American AI Safety Institute. On January 20, 2025, President Trump revoked EO 14110 in an order that rescinded a list of Biden-era executive orders, according to the law firm Wiley.

  • everycrsreport.com statement
    The E.O. directs over 50 federal entities to engage in more than 100 specific actions to implement the guidance set forth across eight overarching policy areas.
  • Time article
    Vice President Kamala Harris announced a sweeping set of U.S. actions, including the establishment of an American AI Safety Institute.

Sources: everycrsreport.com needs a second source · armscontrol.org needs a second source · Time reliable · wiley.law needs a second source

May 2023 – Jun 2023 Center for AI Safety extinction-risk statement

On May 30, 2023, the nonprofit Center for AI Safety (CAIS) released a one-sentence statement: "Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war." CAIS's press release says the signers included Turing Award winners Geoffrey Hinton and Yoshua Bengio and the CEOs of three AI labs: Sam Altman (OpenAI), Demis Hassabis (Google DeepMind) and Dario Amodei (Anthropic). ABC News, as republished by ABC7, said the statement had 350 signers. It also noted the release came about two weeks after Altman told a Senate hearing, "If this technology goes wrong, it can go quite wrong." The statement is a warning about existential risk, not a forecast with a date, so there is no "due date" to check it against.

  • safe.ai statement
    Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.
  • CNN article
    Dozens of AI industry leaders, academics and even some celebrities on Tuesday called for reducing the risk of global annihilation due to artificial intelligence.
  • abc7news.com article
    Among the 350 signatories of the public statement are Sam Altman, the chief executive of OpenAI…

Sources: safe.ai needs a second source · CNN reliable · abc7news.com needs a second source · keyc.com needs a second source

May 2023 Sam Altman testifies to Senate Judiciary subcommittee

On May 16, 2023, OpenAI CEO Sam Altman told a Senate Judiciary subcommittee, "We think that regulatory intervention by governments will be critical to mitigate the risks of increasingly powerful models" (CNN; also in the Tech Policy Press transcript). He suggested "a combination of licensing and testing requirements" for "development and release of AI models above a threshold of capabilities" (CNN). Time reported that he said, "I think if this technology goes wrong, it can go quite wrong, and we want to be vocal about that." CNN reported that he named the use of AI to manipulate voters and target disinformation as among "my areas of greatest concern," because "we're going to face an election next year and these models are getting better." That ties this wave to the thread's later election-related AI disinformation warnings.

  • techpolicy.press statement
    We think it can be a printing press moment. We have to work together to make it so.
  • Time article
    My worst fears are that we—the field, the technology, the industry—cause significant harm to the world. I think that can happen in a lot of different ways.
  • CNN article
    We think that regulatory intervention by governments will be critical to mitigate the risks of increasingly powerful models.

Sources: CNN reliable · Time reliable · techpolicy.press needs a second source

Mar 2023 – Apr 2023 Future of Life Institute "pause" letter

The Future of Life Institute (FLI) is a nonprofit based in Cambridge, Massachusetts (CNBC). Its open letter, dated March 22, 2023 and released after OpenAI's GPT-4 came out earlier that month (CNBC), called on "all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4." It said that "governments should step in and institute a moratorium" if a pause could not be enacted quickly (FLI; CNBC). The letter asked whether society should let machines "flood our information channels with propaganda and untruth," "automate away all the jobs," and "risk loss of control of our civilization" (FLI). Musk, Wozniak and Harari were among more than 1,000 signers when TIME reported it, and Vice named AI researcher Yoshua Bengio as a notable signatory. Vice also reported that some early signatures turned out to be fake. The six-month request came due in late September 2023. Under the headline "The great AI 'pause' that wasn't," Axios reported on Sept. 22, 2023 that "the work is still charging ahead."

  • futureoflife.org statement
    We call on all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4.
  • Time article
    The list of signatories includes the names of more than 1,000 people alongside Musk
  • vice.com article
    some notable signatories turned out to be fake

Sources: futureoflife.org needs a second source · cnbc.com reliable · Time reliable · Axios reliable · siliconangle.com needs a second source · vice.com needs a second source

Mar 2018 Musk at SXSW says AI is "far more dangerous than nukes"

On March 11, 2018, Elon Musk spoke at South by Southwest in Austin, interviewed by Westworld co-creator Jonathan Nolan. He said, "Mark my words. AI is far more dangerous than nukes," and called the arrival of "digital superintelligence" "the single biggest existential crisis that we face." Musk said he is "not normally an advocate of regulation and oversight" but that AI posed "a very serious danger to the public." He proposed a public body with scientific insight and public oversight to make sure AI is developed safely. Inverse noted that this repeated his 2017 National Governors Association warning that AI is a "fundamental risk to the existence of human civilization." The warning had no date attached, so there is no due date to check.

  • medcitynews.com article
    "Mark my words. AI is far more dangerous than nukes," he said.
  • inverse.com article
    "I'm very close to the cutting edge in A.I.," Elon Musk said Sunday while giving a talk at SXSW, "and it scares the hell out of me."

Sources: medcitynews.com needs a second source · inverse.com needs a second source · futurism.com needs a second source · cnbc.com reliable

Jul 2017 Musk tells US governors AI is a "fundamental risk to the existence of human civilization"

On July 15, 2017, Tesla CEO Elon Musk told a bipartisan meeting of U.S. governors that AI regulation is needed because AI is a "fundamental risk to the existence of human civilization" (CBS News/AP; Fortune). He said regulation should come before problems appear, not after: "by the time we are reactive in AI regulation, it's too late" (WBEZ; Fortune has a close variant). The warning was made directly to elected officials. NPR reported that Colorado Gov. John Hickenlooper (D) was looking into how states could respond, and that Arizona Gov. Doug Ducey (R) told Musk he was "surprised by your suggestion to bring regulations before we know exactly what we're dealing with." This was a warning, not a dated forecast, so there is no due date to check.

  • cbsnews.com article
    "Once there is awareness, people will be extremely afraid, as they should be," Musk said.
  • wbez.org article
    "AI is a fundamental existential risk for human civilization, and I don't think people fully appreciate that," Musk said.
  • fortune.com article
    "AI is a fundamental risk to the existence of human civilization, in a way that car accidents, airplane crashes, faulty drugs, or bad food were not."

Sources: cbsnews.com reliable · wbez.org needs a second source · fortune.com reliable · NPR reliable

Data: JSON · CSV · about the data

Related stories