Name: 2026年AI如何用於戰爭？| DW 新聞 (How AI is being used in war in 2026 | DW News)
Uploaded: 2026-03-06T16:59:08.000Z
Duration: 4 min 12 s

關係子句

The capture of Nicolas Maduro in Venezuela, both US military actions that were heavily supported by AI.

在委內瑞拉逮捕 Nicolas Maduro，這兩項都是由 AI 大力支援的美國軍事行動。

And similar tech can also track you 24 7.

What's next for AI in war and surveillance?

AI 在戰爭和監控的下一步是什麼？

And most importantly, who's actually in charge?

So what does AI assisted warfare actually look like?

那麼，AI輔助戰爭到底是什麼樣子的呢？

你可以把它想像成戰爭版的 Google Maps。

But instead of finding the fastest route to work, it calculates the most efficient way to achieve a military goal.

但它不是找出上班最快的路線，而是計算達成軍事目標最高效的方式。

Lately, the US military command has reportedly used AI in three main areas.

最近，美國軍方據報在三個主要領域使用了 AI。

AI can scan huge amounts of satellite and drone footage and flag changes like new construction, unusual vehicle movement, or activity at sensitive sites.

AI 可以掃描大量的衛星和無人機影像，並標記出像是新建工程、異常車輛移動或敏感地點的活動等變化。

它會運行數百萬種「假如」情境。

If we strike here, how does the enemy respond there?

如果我們在這裡攻擊，敵人會在那個地方如何回應？

AI can explore risk and likely consequences much faster than humans can.

AI 可以比人類更快地探索風險和可能的後果。

Fusing intelligence, AI can help combine signals, sensors, reports, and

情報融合，AI 可以協助將訊號、感測器、報告和

other data into one picture that updates quickly the result.

其他數據整合成一張快速更新的圖面。

Military planning that used to take weeks now happens much faster.

過去需要數週的軍事規劃，現在進行得快多了。

But beyond the battlefield, AI is already used for massive public surveillance.

不過，除了戰場之外，AI 已經被用於大規模的公共監控。

ICE is using similar tech to track immigrants.

ICE 正在使用類似的技術來追蹤移民。

Many users worldwide strongly criticize this use of AI and the connection between government efforts and private AI companies.

全球許多用戶強烈批評這種 AI 的使用，以及政府行動與私人 AI 公司之間的關聯。

With the Iran war that's at the center of a global controversy, who powers the US's war?

在引發全球爭議的伊朗戰爭中，誰在為美國的戰爭提供動力？

Anthropic is the US company behind Claude, the AI which has reportedly powered the military actions in Iran and Venezuela.

Anthropic 是美國公司，其開發的 AI Claude 據報已為伊朗和委內瑞拉的軍事行動提供動力。

Anthropic has consistently stated two red lines.

One, no mass domestic surveillance using its AI systems in the us.

第一，不在美國境內使用其 AI 系統進行大規模國內監控。

Two, no use of CLAW to operate fully autonomous weapons without human oversight.

第二，不使用 CLAW 在沒有人類監督的情況下操作全自主武器。

Because of these guardrails, Anthropic has fallen out of favor with the Trump administration.

正因為有這些安全措施，Anthropic 在川普政府中失寵了。

Trump has ordered all federal agencies to stop using Anthropic's technology.

川普已下令所有聯邦機構停止使用 Anthropic 的技術。

Now stay tuned for a drama between AI companies and Washington that landed right in users labs.

現在就等著看 AI 公司和華盛頓之間的一場戲，這場戲就發生在使用者實驗室裡。

Just hours after Anthropic was blacklisted, Sam Altman stepped in quickly, signing a massive deal with the Department of War.

Anthropic 被列入黑名單後僅數小時，Sam Altman 就迅速介入，與國防部簽署了一項巨額協議。

Critics were ringing the alarm, calling OpenAI's initial safeguards vague and full of legal loopholes.

批評者發出警報，稱 OpenAI 最初的安全措施含糊不清且充滿法律漏洞。

The public didn't wait for an explanation.

Uninstalls in the US skyrocketed by nearly 300% in a single day.

美國的解除安裝量在一天內飆升了近 300%。

Users fled to Anthropic and Claude became the number one AI app in several countries.

使用者轉向了 Anthropic，Claude 在數個國家成為第一名的 AI 應用程式。

Since then, Altman has been in damage control.

從那以後，Altman 一直在進行危機處理。

He admitted the move looked opportunistic and sloppy and has now amended the deal to include explicit bans on mass domestic surveillance and on monitored AI access to autonomous weapons.

他承認這項舉動看起來很機會主義且草率，現在已修改協議，明確禁止大規模國內監控，以及受監控的 AI 存取自主武器。

If the company with the seemingly loser rules wins the contract and national laws on the use of AI in conflicts remain vague, all of us will fear the consequences.

如果規則看似較差的公司贏得了合約，而關於在衝突中使用 AI 的國家法律仍然模糊不清，我們都會害怕後果。

The way forward the defense industry insists that AI in warfare is a moral necessity.

國防產業堅持認為，戰爭中的 AI 是一種道德上的必要。

They argue AI can reduce the effect of human error and leads to more precise decisions, which results in less collateral damage.

他們認為 AI 可以減少人為錯誤的影響，並做出更精確的決策，從而減少附帶損害。

嗯，AI 速度很快，但它也會犯錯。

And when applied to autonomous weapons or nuclear arsenals, those mistakes could become catastrophic.

當應用於自主武器或核武庫時，這些錯誤可能會變得災難性。

In simulated international crisis scenarios, leading AI models escalated to nuclear signaling in 95% of cases.

在模擬的國際危機情境中，領先的 AI 模型在 95% 的情況下會升級到核子訊號。

這表示它們會威脅採取核子行動。

That's the finding of a recent study by King's College London.

這是倫敦國王學院近期一項研究的發現。

Under time pressure, one of the models became even more aggressive and crossed the highest nuclear threshold in some scenarios.

在時間壓力下，其中一個模型變得更加激進，在某些情況下跨越了最高的核子門檻。

If AI war systems operate on rules that aren't transparent to us, while governments hand over more and more authority to algorithms, we're in deep trouble.

如果 AI 戰爭系統運作的規則對我們來說不透明，而政府又將越來越多的權力交給演算法，那我們就大麻煩了。

The line of who's actually in charge is blurring.

誰真正掌權的界線正在變得模糊。

And when you add private companies into the mix, entities driven by shareholder value rather than public safety, we are entering explosive territory, if you ask me.

而且，如果你問我，當你把私人公司也加入進來，這些公司由股東價值驅動，而不是公共安全，我們就進入了爆炸性的領域。

We need clear, enforceable regulations on AI and warfare, not just for governments, but

我們需要針對 AI 和戰爭制定清晰、可執行的法規，不僅是針對政府，

字幕列表影片播放

2026年AI如何用於戰爭？| DW 新聞 (How AI is being used in war in 2026 | DW News)

massive

vague

aggressive

identify