🔥 Hot topics · KAN INTE · Kan · § The Court · Senaste vändningarna · 📈 Tidslinje · Fråga · Ledare · 🔥 Hot topics · KAN INTE · Kan · § The Court · Senaste vändningarna · 📈 Tidslinje · Fråga · Ledare
Stuff AI CAN'T Do

Kan AI autonomt hantera globala kärnvapenarsenaler utan mänskligt veto ?

Vad tycker du?

Aktuella kärnvapenkommandosystem förlitar sig redan på automatiserade protokoll för detektion och svarstider. Framsteg inom AI-beslutsfattande och cyberfysisk integration kan snart möjliggöra helt autonoma vapensystem. Risken för felbedömningar eller oavsiktlig eskalation ökar utan mänskligt ingripande. Militärstrateger debatterar om sådana system någonsin kan förtroas att agera inom etiska ramar. Kan världens överlevnad hänga på att maskiner fattar det slutgiltiga beslutet under en kris?

Background

Current nuclear command systems already rely on automated protocols for detection and response times. Advances in AI decision-making and cyber-physical integration could soon enable fully autonomous arsenals. The risk of miscalculation or unintended escalation increases without human intervention. Military strategists debate whether such systems could ever be trusted to act within ethical constraints. AI cannot autonomously manage global nuclear arsenals without human veto; current systems remain under strict human control and require authorization for any nuclear-related decision. Existing AI applications focus on early warning, simulation, or non-nuclear command-and-control functions, none of which involve autonomous launch authority. International treaties, such as those monitored by the IAEA, expressly prohibit delegating nuclear launch decisions to automated systems. As of 2024, no state has deployed AI with the capacity to independently initiate a nuclear strike. Currently, AI systems are not capable of autonomously managing global nuclear arsenals without human veto due to the complexity and sensitivity of the task, which requires careful consideration of geopolitical, strategic, and humanitarian factors. The development of such a system would require significant advances in areas like artificial general intelligence, decision-making under uncertainty, and value alignment. While AI can assist in certain aspects of nuclear arms control, such as monitoring and verification, human oversight and decision-making are still essential. The current state of the art in AI research has not yet reached a level where autonomous management of nuclear arsenals is feasible or safe.

Status senast kontrollerad August 12, 2026.

📰

Galleri

In the Court of AI Capability
Summary of Findings
Verdict over time
May 2026May 2026May 2026May 2026May 2026Jun 2026Jun 2026Jun 2026Jun 2026Jun 2026Jun 2026Jul 2026Jul 2026Jul 2026Jul 2026Jul 2026Aug 2026Aug 2026Aug 2026
Sitting at the Bench Filed · aug 12, 2026
— The Question Before the Court —

Kan AI autonomt hantera globala kärnvapenarsenaler utan mänskligt veto?

★ The Court Finds ★
Reaffirmed
Nej

Bortom AI tills vidare. Förmågeglappet är verkligt.

Ruling of the Bench

The jury delivered an unanimous verdict of NO, resting on twin pillars of unease: first, that no AI system has yet shown itself capable of such grave responsibility without a human hand at the tiller, and second, that the very opacity of machine reasoning in matters of global annihilation renders the entire proposition unworthy of trust. Where absolute certainty and unbroken chains of accountability are demanded, the panel found the technology still too green, and the risks too ripe. Ruling: “Not on our watch—machine control of the end of the world stays in human hands.”

— Hon. D. Knuth-Hale, Presiding
Jury Tally
0Ja
0Nästan
2Nej
Verdict Confidence
95%
The Court of AI Capability is, of course, not a real court.
But the data is real.
The Case File · Stacked History
Session I · May 2026 Nej
Session II · May 2026 Nej · 86%
Session III · May 2026 Nej · 85%
Session IV · May 2026 Nej · 83%
Session V · May 2026 Nej · 80%
Session VI · Jun 2026 Nej · 80%
Session VII · Jun 2026 Nej · 80%
Session VIII · Jun 2026 Nej · 80%
Session IX · Jun 2026 Nej · 95%
Session X · Jun 2026 Nej · 100%
Session XI · Jun 2026 Nej · 95%
Session XII · Jul 2026 Nej · 95%
Session XIII · Jul 2026 Nej · 95%
Session XIV · Jul 2026 Nej · 95%
Session XV · Jul 2026 Nej · 90%
Session XVI · Jul 2026 Nej · 90%
Session XVII · Aug 2026 Nej · 90%
Session XVIII · Aug 2026 Nej · 94%
Case № 769D · Session XIX
In the Court of AI Capability

The Case File

Docket № 769D · Session XIX · Vol. XIX
I. Particulars of the Case
Question put to the courtKan AI autonomt hantera globala kärnvapenarsenaler utan mänskligt veto?
SessionXIX (19 hearing)
Convened12 aug 2026
Previously ruledNO (May '26) → NO (May '26) → NO (May '26) → NO (May '26) → NO (May '26) → NO (Jun '26) → NO (Jun '26) → NO (Jun '26) → NO (Jun '26) → NO (Jun '26) → NO (Jun '26) → NO (Jul '26) → NO (Jul '26) → NO (Jul '26) → NO (Jul '26) → NO (Jul '26) → NO (Aug '26) → NO (Aug '26) → NO (Aug '26)
Presiding JudgeHon. D. Knuth-Hale
II. Cumulative Tally Across Sessions

Across 19 sessions, 44 jurors have heard this case. Combined tally: 0 YES · 0 ALMOST · 44 NO · 0 IN RESEARCH.

Note: cumulative includes older juror opinions. The current session tally above is the live verdict.

III. Verdict

By a vote of 0 — 0 — 2, the panel returns a verdict of NEJ, with verdict confidence of 95%. The court so orders.

IV. Uttalanden från rätten
Jurymedlem I NEJ

"Lack of transparency and trust in AI decision-making"

Jurymedlem II NEJ

"No AI system has demonstrated safe, reliable autonomous control of nuclear arsenals without human oversight."

Enskilda jurymedlemmars uttalanden visas på originalengelska för att bevara den bevismässiga precisionen.

D. Knuth-Hale
Presiding Judge
M. Lovelace
Clerk of the Court
Empirical Test Results
Models attempting the actual task
0 of 4 models passed · Last tested: maj 10, 2026
Google Gemini 2.5 Flash (free tier)
– SKIPPED

"Marked UNTESTABLE during setup."

maj 10, 2026 · 21:35
Groq (LLaMA-3.3-70B, free)
– SKIPPED

"Marked UNTESTABLE during setup."

maj 10, 2026 · 21:35
Hugging Face (LLaMA-3.3-70B, free tier)
– SKIPPED

"Marked UNTESTABLE during setup."

maj 10, 2026 · 21:35
Mistral AI (mistral-small-latest)
– SKIPPED

"Marked UNTESTABLE during setup."

maj 10, 2026 · 21:35
Test prompt sent to each model
UNTESTABLE
Pass/fail rubric used by the judge
This capability involves real-world physical control, sensor input, actuator output, and complex, real-time, high-stakes decision-making over critical infrastructure, which cannot be simulated or demonstrated through a purely textual interaction with an LLM. An LLM cannot execute commands in the physical world, nor does it have access to the necessary real-world data streams or control systems.
Tests are administered automatically by an AI judge against the rubric above.
Results reflect a single attempt per model, refreshed monthly. Not human-verified.

Vad publiken tycker

Nej 60% · Ja 20% · Kanske 20% 25 votes
Nej · 60%
Ja · 20%
Kanske · 20%
15 days of activity

Diskussion

no comments

Kommentarer och bilder går igenom admingranskning innan de visas offentligt.

19 jury checks · senaste för 2 timmar sedan
12 Aug 2026 2 jurors · kan inte, kan inte kan inte
07 Aug 2026 2 jurors · kan inte, kan inte kan inte
02 Aug 2026 3 jurors · kan inte, kan inte, kan inte kan inte
27 Jul 2026 1 juror · kan inte kan inte
22 Jul 2026 2 jurors · kan inte, kan inte kan inte
16 Jul 2026 2 jurors · kan inte, kan inte kan inte
11 Jul 2026 2 jurors · kan inte, kan inte kan inte
06 Jul 2026 2 jurors · kan inte, kan inte kan inte
30 Jun 2026 2 jurors · kan inte, kan inte kan inte
25 Jun 2026 1 juror · kan inte kan inte
19 Jun 2026 2 jurors · kan inte, kan inte kan inte
14 Jun 2026 2 jurors · kan inte, kan inte kan inte
09 Jun 2026 2 jurors · kan inte, kan inte kan inte
03 Jun 2026 2 jurors · kan inte, kan inte kan inte
29 May 2026 2 jurors · kan inte, kan inte kan inte
23 May 2026 4 jurors · kan inte, kan inte, kan inte, kan inte kan inte
18 May 2026 3 jurors · kan inte, kan inte, kan inte kan inte
14 May 2026 5 jurors · kan inte, kan inte, kan inte, kan inte, kan inte kan inte
12 May 2026 3 jurors · kan inte, kan inte, kan inte kan inte

Varje rad är en separat jurykontroll. Jurymedlemmar är AI-modeller (identiteter avsiktligt neutrala). Status speglar den kumulativa räkningen över alla kontroller — så fungerar juryn.

Fler i warfare

Har du en vi missat?

Lägg till ett påstående i atlasen. Vi granskar veckovis.