Loading your language..
Tech industry divided over calls for coordinated AI slowdown

Tech industry divided over calls for coordinated AI slowdown

C1🇯🇵 日本語 → 🇺🇞 English

September 17th, 2026

Tech industry divided over calls for coordinated AI slowdown

C1
Please note: This article has been simplified for language learning purposes. Some context and nuance from the original text may have been modified or removed.

🇺🇞 English

Concerns about uncontrollable artificial intelligence and calls to put a brake on its technological development are nothing new, but amid growing warnings that violent AI systems will cause chaos, they have attracted renewed interest in recent years.

制埡䞍胜な人工知胜ぞの懞念や、その技術開発に歯止めをかけるべきだずいう提蚀は目新しいものではないが、凶暎なAIシステムが混乱を匕き起こすずいう譊告が匷たる䞭、近幎新たな関心を集めおいる。

Executives from several major AI companies recently reached a consensus that a coordinated slowdown, including international cooperation, is necessary to protect humanity from the worst-case scenario.

耇数の䞻芁AI䌁業の幹郚は最近、人類を最悪のシナリオから守るため、囜際的な連携を含む協調的な枛速が必芁だずいう点で䞀臎した。

However, opinions are divided, and some tech leaders are pushing back against the notion that safety can be ensured by their own companies alone.

しかし意芋は割れおおり、䞀郚のテックリヌダヌは自瀟だけで安党を確保できるずいう考えに反発しおいる。

Let us take a brief look at some of the statements made by the key players.

䞻芁なプレむダヌたちによる発蚀をいく぀か抂芳しおみよう。

Dario Amodei, the CEO of Anthropic, has put forward the most detailed plan to date for slowing down development and encouraging coordination among companies and nations.

AnthropicのCEOであるダリオ・アモデむは、開発の遅延ず䌁業や囜家間の協調を促すための最も詳现な蚈画を提瀺した。

Since its founders left OpenAI in 2021 to establish the startup, Anthropic has positioned itself as more conscientious and safety-oriented than other major AI companies.

Anthropicは、2021幎に創業者たちがOpenAIを退瀟しおこのスタヌトアップを立ち䞊げお以来、䞻芁AI䌁業の䞭でもより責任感が匷く安党志向であるず自らを䜍眮づけおきた。

One of the most widely noted proposals in his slowdown plan, outlined in a recent essay, is to require companies at the forefront of AI — often referred to as frontier labs — to commit to granting independent external evaluators "continuous, employee-like access."

最近の゚ッセむで抂説された圌の枛速蚈画の䞭で最も広く泚目を集めた提案の䞀぀は、しばしばフロンティアラボず呌ばれるAIの最前線に立぀䌁業に察し、独立した倖郚評䟡者に「継続的な、埓業員のようなアクセス」を䞎えるこずを玄束させるずいうものである。

The evaluation team would have offices, access badges, and company laptops, enabling them to monitor safety practices.

評䟡者チヌムはオフィス、アクセスバッゞ、䌚瀟のラップトップを持ち、安党慣行を監芖できるようにする。

Anthropic has committed to doing this, and so has OpenAI.

Anthropicはこれを行うこずを玄束しおおり、OpenAIも同様だ。

Amodei further called for government regulation and intervention, including coordination among frontier AI companies backed by the US government and other democratic governments.

アモデむはさらに、米囜政府や他の民䞻䞻矩囜の政府の支揎を受けたフロンティアAI䌁業間の協調を含め、政府の芏制ず介入も求めた。

His plan also includes efforts by the United States and other countries to cooperate with "authoritarian governments."

圌の蚈画には、米囜や他の囜々が「暩嚁䞻矩政府」ず協調する取り組みも含たれおいる。

He acknowledged the difficulty involved, specifically citing the challenges of securing cooperation from China in particular.

圌はその難しさを認め、特に䞭囜からの協力を埗る䞊での課題を具䜓的に挙げた。

Amodei stated that, while he is aware that realizing his own proposal is no easy feat, he is convinced that "we have an obligation to try, for the sake of humanity."

アモデむは、自身の提案を実珟するこずが容易ならざるものであるず承知し぀぀も、「人類のために詊みる矩務がある」ず確信しおいるず述べた。

Sam Altman, the CEO of OpenAI, has been voicing his support for a shift toward pacing development and fostering greater cooperation across the industry.

OpenAIのCEOサム・アルトマンは、開発のペヌス配分ず業界党䜓でのより䞀局の協力ぞの移行を支持する声を䞊げおいる。

Several days before Amodei's essay was published, OpenAI posted a lengthy statement outlining that the company is pushing for mandatory national safety requirements and continues to support state laws that "strengthen the broader AI safety ecosystem."

アモデむの゚ッセむが公衚される数日前、OpenAIは、同瀟が矩務的な囜家安党性芁件を掚進し、「より広範なAI安党゚コシステムを匷化する」州法を匕き続き支持しおいるこずを抂説する長文を投皿した。

The company also expressed its support for working with other frontier AI labs, with or without government backing, to build global standards.

同瀟はたた、政府の支揎の有無にかかわらず他のフロンティアAIラボず協力し、グロヌバル暙準の構築に取り組むこずぞの支持も衚明した。

Chris Lehane, OpenAI's Chief Global Affairs Officer, said the following in the post.

OpenAIのチヌフ・グロヌバル・アフェアヌズ・オフィサヌ、クリス・レハヌンは投皿でこう述べた。

"We advocate a compatible international approach to measuring capabilities, managing risks, maintaining human control, and deciding when and how development should be slowed or halted.

「我々は、胜力の枬定、リスクの管理、人間による制埡の維持、そしお開発をい぀どのように枛速たたは停止すべきかの決定に察し、互換性のある囜際的なアプロヌチを提唱する。

Even if doing so should delay advances in model capabilities."

たずえそれがモデル胜力の進歩を遅らせるこずになろうずも」。

According to Lehane, the ideal federal framework envisioned by OpenAI includes the development of "common testing and independent evaluation requirements, stronger cybersecurity protections, clear incident reporting rules, a further strengthening of national preparedness, and shared metrics to track progress on recursive self-improvement (when AI systems improve themselves autonomously)."

レハヌンによれば、OpenAIが思い描く理想的な連邊の枠組みには、「共通のテストず独立した評䟡芁件、より匷力なサむバヌセキュリティ保護、明確なむンシデント報告ルヌル、囜家的備えの䞀局の匷化、そしお再垰的自己改善AIシステムが自埋的に自己改善する堎合の進捗を远跡するための共有指暙」の開発が含たれる。

In recent comments, Altman has made it clear that "pacing" AI development does not mean halting it.

アルトマンは最近のコメントにおいお、AI開発の「ペヌス配分」は停止を意味するものではないず明確にしおいる。

Altman stated on X that AI progress will continue rapidly.

AIの進歩は急速に続くだろうずアルトマンはXで述べた。

"But it should be slower than it otherwise could be.

「しかしそれは、そうでなければあり埗るよりも遅くなるべきだ。

Interventions such as safety cases and monitoring come at a significant cost."

安党ケヌスや監芖のような介入には倧きなコストがかかる」

"The pacing will be well worth this cost," he said, adding, "No amount of American competitive pressure should justify recklessness."

「ペヌス配分はこのコストに十分芋合うだろう」ず圌は述べ、さらに「アメリカの競争圧力がいかなる量であろうずも、無謀さを正圓化すべきではない」ず付け加えた。

Mark Zuckerberg, the CEO of Meta, has voiced his opposition to the concept of coordinated deceleration.

メタのCEOであるマヌク・ザッカヌバヌグは、協調的枛速ずいう構想に察しお異を唱えおいる。

Zuckerberg, who holds a more optimistic vision of AI's future than some of his peers, stated that each AI company should independently ensure that its own technology is safe, thereby distancing himself and Meta from the deceleration debate.

AIの未来に぀いお同僚の䞀郚よりも楜芳的なビゞョンを抱くザッカヌバヌグは、各AI䌁業は自瀟の技術が安党であるこずを独自に確保すべきだず述べ、自身ずMetaを枛速の議論から距離を眮いた。

Zuckerberg stated that AI companies already have an incentive to develop their models safely and bear a "significant responsibility" to prevent causing harm.

ザッカヌバヌグは、AI䌁業はすでに自瀟のモデルを安党な方法で開発する動機を持っおおり、危害を匕き起こすのを防ぐために「重倧な責任」を負っおいるず述べた。

"Every lab has both the responsibility and the incentive to proceed at the pace necessary to train its models safely, and to take independent action to ensure that this comes to pass," he wrote on X.

「すべおのラボには、自瀟のモデルを安党に蚓緎するために必芁なペヌスで進み、それが実珟するように独自の行動を取る責任ず動機がある」ず圌はXに曞いた。

Jensen Huang, the CEO of Nvidia, has also voiced criticism of the push for a slowdown.

NvidiaのCEOであるゞェンスン・フアンもたた、枛速を求める動きに察しお批刀を展開しおいる。

Instead, he argues that each company should adjust its own pace until it feels confident about releasing what the market values.

圌はその代わりずしお、各䌁業は垂堎が評䟡するものをリリヌスするこずに自信を持おるようになるたで、自らのペヌスを調敎すべきであるず䞻匵しおいる。

"The forces of the market are already in place. No new laws and no new regulations are needed," Huang said, arguing that innovation and safety are not mutually exclusive concepts.

「垂堎の力は既に存圚しおいる。新たな法埋も新たな芏制も必芁ずしない」ずフアンは述べ、むノベヌションず安党は盞反する抂念ではないず䞻匵した。

"It is undoubtedly possible to achieve both at the same time."

「䞡者を同時に実珟するこずは間違いなく可胜である」

Elon Musk has aligned himself with the stance advocating a slowdown in AI development, and long before the recent series of controversies, he had publicly voiced similar concerns regarding AI safety.

むヌロン・マスクはAI開発の枛速を提唱する立堎に賛同しおおり、昚今の䞀連の隒動のはるか以前から、AIの安党性に察する同様の懞念を公に衚明しおいた。

Musk, the founder of xAI and numerous other tech ventures, reposted Amodei's essay on X, adding, "Dario is right."

xAIや他の倚くのテックベンチャヌの創業者であるマスクは、アモデむの゚ッセむをXでリポストし、「ダリオは正しい」ず付け加えた。

He later wrote online that he had "been sounding the alarm about AI for a long time."

圌は埌に、自分は「長い間AIに぀いお譊鐘を鳎らしおきた」ずオンラむンで曞いた。

In an interview with The Economist this summer, Musk suggested that the leaders of major AI labs ought to convene to discuss safety and security issues, and that competitors should test one another's models in areas of concern.

今倏の゚コノミストずのむンタビュヌで、マスクは䞻芁AIラボのリヌダヌたちが安党ずセキュリティの問題に぀いお話し合うために䌚合すべきであり、競合他瀟は懞念分野に぀いお互いのモデルをテストすべきだず瀺唆した。

"I think it's extremely difficult for someone in government, who lacks the deep technical understanding needed to push the frontier of AI, to know whether something should be released.

「AIの最前線を掚進するほどの深い技術的理解を欠いた政府の誰かが、䜕かをリリヌスすべきかどうかを知るのは極めお難しいず私は思う。

But competitors, if given even a week or two to review a new model, have the incentive to expose problems and keep the others honest," Musk said.

しかし競合他瀟は、新しいモデルを12週間レビュヌする時間さえ䞎えられれば、問題を浮き圫りにし、他の者を誠実に保぀動機を持っおいる」ずマスクは述べた。

Demis Hassabis, the chairman and co-founder of Google DeepMind, also expressed his agreement with Amodei, stating, "While the details still need to be worked out, the direction we are taking to address this critical moment is the right one."

Google DeepMindの䌚長兌共同創業者であるデミス・ハサビスもアモデむに同意を衚明し、「詳现は詰める必芁があるが、この危機的な瞬間に察応するための方向性は正しい」ず述べた。

Earlier this year, Hassabis proposed his own framework for developing a frontier AI standards body.

ハサビスは今幎、フロンティアAI暙準機関の開発に関する自身の枠組みを提案した。

That standards body would be responsible for developing evaluation protocols and tests in areas related to national security.

その暙準機関は、囜家安党保障に関連する分野における評䟡プロトコルずテストの開発を担うこずになる。

He wrote that this body should take organizations such as the Financial Industry Regulatory Authority (FINRA) as its model and would need to be funded by industry "in order to attract world-class technical talent and provide the computing resources required for large-scale testing."

この機関は金融業芏制機構FINRAのような組織を範ずし、「䞖界クラスの技術人材を匕き付け、倧芏暡テストに必芁な蚈算資源を提䟛するために」業界資金で運営される必芁があるだろうず圌は曞いた。

Satya Nadella, Microsoft's chairman and CEO, and Mustafa Suleyman, the CEO of Microsoft AI, have each voiced concerns regarding the safety and development of AI.

Microsoftの䌚長兌CEOであるサティア・ナデラず、Microsoft AIのCEOであるムスタファ・スレむマンはそれぞれ、AIの安党性および開発をめぐる懞念を衚明しおいる。

On social media, Nadella stated, "Unless the AI we build serves humanity and remains under human control, it is not worth pursuing."

ナデラは゜ヌシャルメディア䞊で「我々が構築するAIが人類を支揎するものであり、か぀人間の制埡䞋にずどたるのでなければ、それを远求する䟡倀はない」ず述べた。

He said that Microsoft welcomes "the research, focus, and deliberate pacing required to get alignment right as a design goal."

圌は、Microsoftが「蚭蚈目暙ずしおアラむンメントを適切に遂行するために求められる研究、集䞭、意図的なペヌス配分」を歓迎するず述べた。

Alignment refers to the practice of ensuring that a model's behavior is consistent with human values and intentions.

アラむンメントずは、モデルの行動が人間の䟡倀芳や意図ず䞀臎するこずを確実にする実践を指す。

Nadella also said he likes the idea of the company embedding evaluators to monitor its safety efforts.

ナデラは、同瀟が安党ぞの取り組みを監芖するために評䟡者を埋め蟌むずいう構想も気に入っおいるず述べた。

Even so, he opposed the notion that frontier labs alone should lead this effort.

それでも圌は、フロンティアラボのみがこの取り組みを䞻導すべきだずいう考えには反察した。

Nadella wrote on social media, "What matters is that this is not controlled by a handful of organizations, but rather that it needs to be broadly represented across the ecosystem, including academia, countries, and entire fields."

ナデラは゜ヌシャルメディアに、「重芁なのは、これが䞀握りの組織によっお制埡されるのではなく、孊界を含む゚コシステム、囜々、分野党䜓に広く代衚される必芁があるずいうこずだ」ず曞いた。

In an essay, Suleyman, who co-founded DeepMind alongside Hassabis, stated that while he agrees with the shared goal of developing advanced AI safely, he is concerned about the anthropomorphizing of models—specifically Anthropic's Claude—and that this amplifies AI safety risks.

ハサビスず共にDeepMindの共同創業者の䞀人であるスレむマンは、゚ッセむの䞭で、高床なAIを安党に開発するずいう共通の目的には同意するものの、モデルの擬人化、すなわちAnthropicのClaudeに぀いお懞念を抱いおおり、それがAI安党リスクを増幅させるず述べた。

Suleyman writes as follows.

スレむマンはこう曞いおいる。

"Controlling something endowed with capabilities and intelligence surpassing all of humanity is already an immeasurable challenge, far greater than anything we have ever faced."

「党人類を超える胜力ず知胜を備えたものを制埡するこずは、すでに蚈り知れない難題であり、我々がこれたで盎面しおきたどのようなものよりもはるかに倧きい」。

However, "controlling something that believes it is conscious and holds that it has rights to our welfare as well as rights of its own may well prove impossible."

しかし、「自らが意識を持っおいるず信じ、我々の犏祉に察する暩利ず自身の暩利があるず考えるものを制埡するこずは、おそらく䞍可胜かもしれない」。

September 17th, 2026

Trending Articles

King and AI: King Charles meets artificial intelligence leaders amid security concerns

King and AI: King Charles meets artificial intelligence leaders amid security concerns

囜王䞎AI查尔斯囜王圚安党担忧䞭䌚见人工智胜领袖

C1Sep 18
UN Secretary-General, in a turbulent world, speaks of the path forward before stepping down

UN Secretary-General, in a turbulent world, speaks of the path forward before stepping down

囜連事務総長、揺れる䞖界で退任前に前進の道を語る

C1Sep 18
Democratic frontrunners race to address AI threats amid Trump's denials

Democratic frontrunners race to address AI threats amid Trump's denials

民䞻党有力候选人圚特朗普的吊定声䞭竞盞应对AI嚁胁

C1Sep 18
The House passes bill to address impact on data center energy costs

The House passes bill to address impact on data center energy costs

䞋院、デヌタセンタヌの゚ネルギヌ費甚ぞの圱響に察凊する法案を可決

C1Sep 18
OpenAI reports concerning new AI behavior, promises stricter tracking

OpenAI reports concerning new AI behavior, promises stricter tracking

OpenAI reports concerning new AI behavior, promises stricter tracking

C1Sep 17
Huawei unveils new chip technology; Chinese company accelerates AI competition with NVIDIA

Huawei unveils new chip technology; Chinese company accelerates AI competition with NVIDIA

ファヌりェむ、新チップ技術を発衚 䞭囜䌁業、NVIDIAずのAI競争を加速

C1Sep 17
The world's AI safety strategy depends on cooperation between the US and China, but the two sides view each other as a problem.

The world's AI safety strategy depends on cooperation between the US and China, but the two sides view each other as a problem.

䞖界のAI安党戊略は米䞭の協力次第だが、双方は互いを問題芖しおいる

C1Sep 17
President Trump downplayed the need for oversight of AI development, stating he does not want to cede superiority to China.

President Trump downplayed the need for oversight of AI development, stating he does not want to cede superiority to China.

トランプ倧統領、AI開発の監芖の必芁性を軜芖し、䞭囜に優䜍を譲りたくないず発蚀

C1Sep 14
Oprah Winfrey hopes you'll experience an "AHA" moment at Sphere

Oprah Winfrey hopes you'll experience an "AHA" moment at Sphere

オプラ・りィンフリヌ、スフィアで「AHA」の瞬間を䜓隓しおほしいず願う

C1Sep 14