Physical intervention (literally unplugging data centers) may be required once AI agents begin communicating in proprietary, non-human-decipherable languages.

Physical intervention (literally unplugging data centers) may be required once AI agents begin communicating in proprietary, non-human-decipherable languages. | Multi-agent emergent-communication research

Key catalyst: Multi-agent emergent-communication research

Watch events: Multi-agent emergent communication research; AI governance legislation; kill-switch policy proposals.

Multi-agent systems research (OpenAI, Anthropic, DeepMind) shows emergent coordination patterns; 'non-human language' between agents is observable in multi-agent RL.

Adverse (2)

{
  "nia": false,
  "qty": "DC-level kill switch",
  "mode": "THESIS",
  "role": "Guest-CEO",
  "context": "Schmidt's ultimate-safeguard framing; extends his long-running US-China AI-race concern into concrete kill-switch scenario.",
  "to_year": 2030,
  "conv_cues": "may become necessary; ex-Google CEO gravitas",
  "direction": "HAPPEN",
  "from_year": 2026,
  "timeframe": "this decade",
  "conv_level": "HIGH",
  "milestones": [
    {
      "kind": "llm_pre_event",
      "label": "Peer-reviewed paper shows 2 LLM agents develop shared lexicon over 541 objects in 4 rounds",
      "notes": "HIT — earliest empirical evidence that emergent LLM-agent languages are realistic, foundational to the 'unplug datacenters' thesis.",
      "source": "https://openreview.net/forum?id=zy06mHNoO2 — Emergence of Machine Language in LLM-based Agent Communication",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -8,
      "source_id": null,
      "confidence": 0.95,
      "source_url": "https://openreview.net/forum?id=zy06mHNoO2",
      "expected_date": "2025-11-14",
      "observed_date": "2025-12-15",
      "research_origin": "deep_research",
      "expected_date_range": {
        "to": "2026-04-30",
        "from": "2025-06-01"
      },
      "measurement_criterion": "Peer-reviewed publication (OpenReview / arXiv accepted) demonstrates two LLM agents independently converging on a shared 541-object vocabulary in only 4 rounds of communication"
    },
    {
      "kind": "llm_pre_event",
      "label": "AAAI 2026 Bridge Program convenes on multi-agent LLM emergent communication",
      "source": "https://arxiv.org/html/2511.17332v2 — Agentifying Agentic AI / WMAC 2026 / AAAI 2026 Bridge",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -7,
      "source_id": null,
      "confidence": 0.92,
      "source_url": "https://arxiv.org/html/2511.17332v2",
      "expected_date": "2026-02-15",
      "observed_date": "2026-02-15",
      "research_origin": "deep_research",
      "measurement_criterion": "AAAI 2026 hosts dedicated bridge / workshop on Advancing LLM-Based Multi-Agent Collaboration with research on emergent communication"
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q1 window check-in (25%)",
      "status": "pending",
      "weight": 0.05,
      "ordinal": -6,
      "source_id": null,
      "expected_date": "2026-10-24",
      "observed_date": null
    },
    {
      "kind": "llm_pre_event",
      "label": "First documented inter-agent message corpus that human evaluators cannot reliably interpret",
      "notes": "Direct measurement of the prediction's named precondition (proprietary non-human-decipherable language).",
      "source": "Alignment Forum / Anthropic / OpenReview research",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -5,
      "source_id": null,
      "confidence": 0.55,
      "expected_date": "2027-06-16",
      "research_origin": "training",
      "expected_date_range": {
        "to": "2028-06-30",
        "from": "2026-06-01"
      },
      "measurement_criterion": "Published study reports >70% inter-agent communication tokens are non-decipherable by human annotators with full context, in production-class multi-agent system"
    },
    {
      "kind": "llm_pre_event",
      "label": "AI lab publicly proposes 'kill switch / circuit breaker' protocol for emergent-comm-MAS",
      "notes": "Operationalizes 'unplug datacenters' from rhetoric to safety policy.",
      "source": "https://alignment.anthropic.com/ — Anthropic Alignment Science Blog",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -4,
      "source_id": null,
      "confidence": 0.55,
      "source_url": "https://alignment.anthropic.com/",
      "expected_date": "2027-06-16",
      "research_origin": "deep_research",
      "expected_date_range": {
        "to": "2028-06-30",
        "from": "2026-06-01"
      },
      "measurement_criterion": "OpenAI / Anthropic / DeepMind publishes 
... (truncated)