← Cockpit
CYB_022predictionAILiteLLM-supply-chain-vulnerability

Supply-chain attacks targeting widely utilized open-source AI packages — exemplified by documented attacks on the LiteLLM framework — represent extreme systemic risk: a single compromised foundational code repository instantly and autonomously infects ...

Predictor: Alex Wissner-Gross

Prior probability
78.0%
Current probability
75.9%
evolves via intake + LBP
Conviction
5/5
Signal quality
B
Resolution
in_progress
Window
2026-01-01 – 2028-11-30
Edges in / out
2 / 0
Tickers exposed
13

Prediction text

Supply-chain attacks targeting widely utilized open-source AI packages — exemplified by documented attacks on the LiteLLM framework — represent extreme systemic risk: a single compromised foundational code repository instantly and autonomously infects thousands of downstream enterprise agent networks at machine speed, bypassing traditional firewall perimeters entirely. | Next major AI-package supply chain CVE

Key catalyst: Next major AI-package supply chain CVE

Watch events: PyPI/npm AI-package attack incidents; enterprise SBOM mandates

Resolution evidence

Status: in_progress

LiteLLM CVE-2024/2025 disclosures documented; PyPI / npm AI-package supply-chain attacks increasing 2024-2026. Snyk, Socket.dev reporting sharp rises.

Predictor: Alex Wissner-Gross

κ + Brier as of 2026-05-22
κ (discount)
0.844
Brier
0.0341
excellent
Hits / Misses
6 / 1
of 11 resolved
Hit rate
54.5%
Calibration plot (stated vs observed)

Evidence about this node from Alex Wissner-Gross is multiplied by κ in /api/intake. Lower κ = less weight; floors at 0.10 (effectively silenced) and caps at 1.00 (full weight).

Reference class: ai_catastrophic_misuse_1y

Linked

Frontier AI used in successful catastrophic-class (bio/cyber/chem) attack within 1y of capability claim

Base rate
1/20 historical
Inside weight
Outside weight
no pull
inside 75.9% → blend 75.9% 0.0pp)

Tetlock-style outside view: at TRF=1 (just predicted), outside view dominates (w_in=0.3). At TRF=0 (deadline), inside view dominates (w_in=1.0). The blend regularizes overconfident inside views toward the historical base rate.

Probability over time

4 prob_history rows
0%25%50%75%100%prior 78%2026-04-302026-05-212026-05-24
intake v2milestone miss sweeplbp propagationreference class assignedlegacy v1prior_prob (analyst seed)current = 75.9%

Milestone chain

Pre-event signals (upstream prereqs + window checkpoints) → resolution event → downstream cascades. Status/dates update from linked nodes; re-derive nightly via scripts/ops/derive_milestones.py.
Leading chain: 3 fired ✓ · 5 pending
  1. 2026-03-01hitCVE-2026-33634 issued for upstream Trivy / TeamPCP campaign
    How: MITRE/NVD assigns CVE with CVSS ≥9.0 for AI-toolchain supply-chain attack with documented downstream cascade
    Source: https://www.kaspersky.com/blog/critical-supply-chain-attack-trivy-litellm-checkmarx-teampcp/55510/conf 95%
    Notes: HIT — CVE-2026-33634 with CVSS 9.4. Trivy → LiteLLM cascade was attack mechanism.
  2. 2026-03-24hitLiteLLM PyPI supply-chain compromise (versions 1.82.7, 1.82.8)
    How: Documented compromise of widely-used AI package on PyPI/npm with confirmed downstream propagation
    Source: https://docs.litellm.ai/blog/security-update-march-2026 — LiteLLM 1.82.7/1.82.8 compromised March 24 2026conf 99%
    Notes: HIT — exactly the event Wissner-Gross referenced. LiteLLM downloaded 3.4M times/day, ≥20K downstream repos potentially exposed.
  3. 2026-03-24hitThree-stage payload hits credential / Kubernetes / backdoor at scale
    How: Compromised package shown to deploy credential harvester targeting ≥50 secret types + lateral-movement toolkit + persistent backdoor
    Source: https://www.helpnetsecurity.com/2026/03/25/teampcp-supply-chain-attacks/conf 99%
  4. 2026-07-21pendingQ1 window check-in (25%)
  5. 2026-06-01 → 2026-12-31pendingSecond wave of AI-package supply-chain CVE during 2026
    How: ≥1 additional widely-deployed AI/agent package compromised at PyPI/npm scale during 2026, after LiteLLM
    Source: TeamPCP campaign tracked as 'Phase 09' — implies further phases pending; Wiz/Snyk/Kaspersky monitoringconf 75%
  6. 2027-02-07pendingQ2 window check-in (50%)
  7. 2026-09-01 → 2027-12-31pendingEnterprise breach attributable to AI-package compromise reported in SEC 8-K
    How: Public-company 8-K filing discloses material breach attributable to compromised open-source AI package
    Source: SEC EDGAR, security-incident filingsconf 55%
    Notes: Cascade — operationalizes Wissner-Gross's 'systemic risk' framing as a measurable enterprise impact event.
  8. 2027-08-27pendingQ3 window check-in (75%)

No downstream cascades — this prediction is a leaf in the dependency graph.

What if this resolves?

Clamp this prediction TRUE or FALSE and run a counterfactual Gibbs sample. Surfaces the predictions whose marginals shift most under that assumption.
(live posterior: 76%)

Click a button to clamp this prediction and run a Gibbs sample. Returns the predictions whose marginals shift most. ~30s per run; ideal for stress-testing "if X resolves, what else moves?"

Evidence chain

Every probability update with full Bayesian provenance — chronological, latest first
LBP2026-05-24T02:00:02Z75.9%-3.5pp
Network propagation: 79.4% → 75.9%
4-iter LBP, residual 0.01000 · damping 0.5, w_intrinsic 0.5 · method lbp_v3 · run 806b02f8
intake_event_update2026-05-21T23:15:16Z79.4%+6.1pp
intake:7afeeb9a-f217-4dd2-b910-24ff14bdfc39 bayesian_v2 inside=0.794 blend=0.794 LLR=0.342 κ=0.84 no_blend
Raw metadata
{
  "trf": 0.8675103990988756,
  "kappa": 0.8438,
  "base_rate": null,
  "predictor": "Alex Wissner-Gross",
  "total_llr": 0.4054651081081644,
  "bayesian_v2": true,
  "prior_logit": 1.009422591843683,
  "bayes_factor": "1.4:1 favoring",
  "blend_reason": "no reference_class linked",
  "inside_prior": 0.7329071343302511,
  "kappa_source": "predictor_table",
  "blend_applied": false,
  "contributions": [
    {
      "llr": 0.4054651081081644,
      "kappa": 0.8438,
      "label": "Industry concern about AI supply-chain / packaging attacks now mainstream enough to motivate dedicated cyber-AI initiati",
      "adjusted_llr": 0.3421314582216691
    }
  ],
  "evidence_kind": "intake_event_update",
  "inside_source": "history_v2",
  "inside_weight": 1,
  "outside_weight": 0,
  "posterior_prob": 0.7943835802353901,
  "evidence_origin": "daily_intake",
  "llm_suggestions": [
    {
      "polarity": "corroborates",
      "status_change": "unchanged",
      "evidence_strength": "weak",
      "delta_prob_suggestion": 0.03
    }
  ],
  "posterior_logit": 1.351554050065352,
  "predictor_brier": 0.03413,
  "evidence_doc_ids": [],
  "inside_posterior": 0.7943835802353901,
  "blended_posterior": 0.7943835802353901,
  "reference_class_id": null,
  "total_adjusted_llr": 0.3421314582216691,
  "predictor_n_resolved": 11
}
LBP2026-04-30T16:39:51Z73.3%-1.7pp
Network propagation: 75.0% → 73.3%
5-iter LBP, residual 0.00825 · damping 0.5, w_intrinsic 0.5 · method lbp_v2 · run 0c8a4ea3
LBP2026-04-30T02:18:57Z75.0%-3.0pp
Network propagation: 78.0% → 75.0%
5-iter LBP, residual 0.00825 · damping 0.5, w_intrinsic 0.5 · method lbp_v1 · run 592311ef

Network propagation neighbors

Top edges sorted by latest LBP cross-impact
All propagation →

Top incoming (parents)

Edges that influence THIS node's belief

KindNodeTheir probP(c|s=T)P(c|s=F)Δ implied
killerTK11
Autonomous Regulatory Block (Level 4 Halt)
10.0%0.0500.780-0.052
killerTK06
China-Taiwan Military Conflict
8.0%0.0500.780-0.037

Top outgoing (children)

Predictions THIS node influences

No outgoing edges.

Ticker exposure

13 ticker(s) linked

Beneficiaries (13)

AIBBAIGTLBNVDASOUNIBMMETAMSFTSHOPAMZNORCLGOOGLPLTR

Prerequisites (2)

Predictions that must hit first
TypePredTitleDomainLag
killerTK11Autonomous Regulatory Block (Level 4 Halt)
killerTK06China-Taiwan Military Conflict

Dependents (0)

Predictions enabled by this
TypePredTitleDomainLag
No dependents

Validations (1)

Resolution events
Observed atStatusByNotes
2026-04-29partialthesis_timeline_v1.0_importLiteLLM CVE-2024/2025 disclosures documented; PyPI / npm AI-package supply-chain attacks increasing 2024-2026. Snyk, Socket.dev reporting sharp rises.

Linked documents (10)

Auto-generated by cosine similarity from Polymarket / Manifold / EDGAR / GDELT

Raw metadata

From Thesis_Timeline_v1.0_FINAL workbook
{
  "nia": false,
  "mode": "FORECAST",
  "role": "Cited-Other",
  "context": "Specific vector extending 231_017 (Wissner-Gross: major supply chain attack from untrusted open-weight models is expected). LiteLLM is named vulnerable component.",
  "to_year": 2028,
  "conv_cues": "specific named framework vulnerability; severe framing",
  "direction": "HAPPEN",
  "from_year": 2026,
  "timeframe": "2026-2028",
  "conv_level": "HIGH",
  "milestones": [
    {
      "kind": "llm_pre_event",
      "label": "CVE-2026-33634 issued for upstream Trivy / TeamPCP campaign",
      "notes": "HIT — CVE-2026-33634 with CVSS 9.4. Trivy → LiteLLM cascade was attack mechanism.",
      "source": "https://www.kaspersky.com/blog/critical-supply-chain-attack-trivy-litellm-checkmarx-teampcp/55510/",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -8,
      "source_id": null,
      "confidence": 0.95,
      "source_url": "https://www.kaspersky.com/blog/critical-supply-chain-attack-trivy-litellm-checkmarx-teampcp/55510/",
      "expected_date": "2026-03-01",
      "observed_date": "2026-03-01",
      "research_origin": "deep_research",
      "measurement_criterion": "MITRE/NVD assigns CVE with CVSS ≥9.0 for AI-toolchain supply-chain attack with documented downstream cascade"
    },
    {
      "kind": "llm_pre_event",
      "label": "LiteLLM PyPI supply-chain compromise (versions 1.82.7, 1.82.8)",
      "notes": "HIT — exactly the event Wissner-Gross referenced. LiteLLM downloaded 3.4M times/day, ≥20K downstream repos potentially exposed.",
      "source": "https://docs.litellm.ai/blog/security-update-march-2026 — LiteLLM 1.82.7/1.82.8 compromised March 24 2026",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -7,
      "source_id": null,
      "confidence": 0.99,
      "source_url": "https://docs.litellm.ai/blog/security-update-march-2026",
      "expected_date": "2026-03-24",
      "observed_date": "2026-03-24",
      "research_origin": "deep_research",
      "measurement_criterion": "Documented compromise of widely-used AI package on PyPI/npm with confirmed downstream propagation"
    },
    {
      "kind": "llm_pre_event",
      "label": "Three-stage payload hits credential / Kubernetes / backdoor at scale",
      "source": "https://www.helpnetsecurity.com/2026/03/25/teampcp-supply-chain-attacks/",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -6,
      "source_id": null,
      "confidence": 0.99,
      "source_url": "https://www.helpnetsecurity.com/2026/03/25/teampcp-supply-chain-attacks/",
      "expected_date": "2026-03-24",
      "observed_date": "2026-03-24",
      "research_origin": "deep_research",
      "measurement_criterion": "Compromised package shown to deploy credential harvester targeting ≥50 secret types + lateral-movement toolkit + persistent backdoor"
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q1 window check-in (25%)",
      "status": "pending",
      "weight": 0.05,
      "ordinal": -5,
      "source_id": null,
      "expected_date": "2026-07-21",
      "observed_date": null
    },
    {
      "kind": "llm_pre_event",
      "label": "Second wave of AI-package supply-chain CVE during 2026",
      "source": "TeamPCP campaign tracked as 'Phase 09' — implies further phases pending; Wiz/Snyk/Kaspersky monitoring",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -4,
      "source_id": null,
      "confidence": 0.75,
      "source_url": "https://snyk.io/blog/poisoned-security-scanner-backdooring-litellm/",
      "expected_date": "2026-09-15",
      "research_origin": "deep_research",
      "expected_date_range": {
        "to": "2026-12-31",
        "from": "2026-06-01"
      },
      "measurement_criterion": "≥1 additional widely-deployed AI/agent package compromised at PyPI/npm scale during 2026, after LiteLLM"
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q2 window check-in (50%)",
      "status": "pending",
      "weight": 0.05,
      
... (truncated)