Due Diligence Checklist for Trustees Evaluating AI and Early-Stage Tech Investments

Why trustees must run rigorous AI startup due diligence in 2026

Trustees managing trust exposure to venture-stage technology face a new set of high-stakes questions in 2026. Rapid shifts in AI compute demand, concentrated infrastructure risk and headline losses at early AI companies mean that a few bad allocations can erode decades of trust wealth. If you are weighing an allocation to an AI startup or holding private shares inside a trust, this checklist helps you answer the practical questions trustees need now: Does the company have product-market fit? Is the fundraising runway real? Is the IP enforceable? What are credible exit scenarios?

Fast context: two 2025–2026 signals trustees cannot ignore

Recent reporting (January 2026) flagged two trends that changed the underwriting calculus for AI investments. First, multiple early AI platform companies — notably a high-profile firm publicly reported as struggling with product strategy and fundraising — showed how quickly teams and talent move away when a clear commercial plan is lacking. That case underlines the risk of product ambiguity and talent flight.

Second, late-2025 reporting on grid and infrastructure planning (including a proposed PJM auction driven by AI demand) highlights systemic exposure: large-scale AI workloads now compete for energy and data-center capacity, creating operational and margin risks for startups that rely on predictable, cheap compute.

How trustees should use this checklist

This is a working due diligence tool optimized for trustees, professional fiduciaries and small-business owners who hold private securities in trusts. Use it to:

• Score and compare AI startup opportunities across consistent criteria.
• Identify governance protections and contract terms to request before committing capital.
• Limit trust exposure with quantitative caps and conditional milestones.

Structure: three-stage checklist

The checklist is organized into three actionable stages: (A) Quick screening (10–20 minutes), (B) Deep diligence (2–4 weeks), and (C) Governance & risk mitigation (contractual terms trustees must negotiate). Each item includes concrete evidence to request and how to score it.

A. Quick screening — immediate red flags (1–2 items you must check right away)

Use this stage to eliminate opportunities that pose unacceptable trust exposure.

1. Clear product definition and revenue model (Score 0–3)
   • Evidence to request: 1-page product one-pager, GTM slides, recent customer contracts, revenue run-rate and ARR if applicable.
   • Red flags: no paying customers or pilot revenue after 12+ months; conflicting descriptions of the product; changing target markets every quarter (see Thinking Machines case where product strategy reportedly lacked focus).
2. Fundraising runway and investor appetite (Score 0–3)
   • Evidence: latest cap table, cash balance, monthly burn, committed deals or signed LOIs for the current raise, recent valuation history and lead investor term sheet.
   • Red flags: failure to secure lead investor after multiple pitches; undisclosed side letters; payroll or vendor arrears; aggressive bridge financing with punitive terms.

B. Deep diligence — detailed evaluation across nine domains

Score each domain 0–5, then apply weightings (example weighting below). Ask for documentary evidence and independent verification where possible.

1. Product-market fit & commercial traction (Weight 20%)

• Metrics to request: ARR/NRR, churn, customer concentration, pilot conversion rate, average contract size, LTV/CAC, usage patterns, customer references (3+), and letters of intent.
• Qualitative checks: Are customers paying meaningful amounts? Are there multi-year contracts or only POCs? Are product roadmaps aligned with customer pain?
• Red flags: Heavy reliance on free pilots, no conversion after 6–12 months, single-customer concentration >30% of revenue.

2. Team, retention & hiring risk (Weight 15%)

• What to verify: Founders' track records, key hires' employment agreements, non-compete and IP assignment confirmations, recent departures and exit interviews.
• Red flags: Mass employee conversations with competitors (as reported in Jan 2026 about employees in talks with larger AI platforms), founder infighting, missing IP assignments.

3. Fundraising runway & capitalization (Weight 15%)

• Documents: Latest cap table (post-money and fully diluted), convertible notes and SAFEs, investor term sheets, most recent investor update, cashflow forecast for 12–24 months.
• Metrics: Months of runway at current burn, milestone-based funding triggers, committed vs. expected capital.
• Red flags: Understated burn projections, reliance on indefinite bridges, aggressive pro-rata promises with no lead investor.

4. Technology & infrastructure risk (Weight 15%)

• Key checks: Architecture diagrams, dependency map (data, cloud, GPUs, custom silicon), MLOps reproducibility (can models be retrained), disaster recovery and capacity contracts.
• Energy and capacity: Ask how compute is procured and hedged. With grid constraints and auctions forming for AI capacity (PJM developments in late 2025), startups with unpredictable compute needs face margin and continuity risk.
• Red flags: Single data-center dependency, no long-term cloud commitments, reliance on spot instances with high cost volatility.

5. Intellectual property and data provenance (Weight 12%)

• IP evidence: Patent filings, assignment agreements, trade-secret policies, employee IP assignment confirmations, third-party licensing agreements for models or data, data usage licenses and provenance documentation.
• AI-specific risks: Training data rights, upstream license scopes (open-source model licenses can impose restrictions), and policies for derivative works. Model theft and dataset claims are increasing in 2025–26; verify chain-of-title.
• Red flags: No documented data licenses for training sets, ambiguous third-party model use, absent employee IP assignments.

6. Regulatory, privacy and export risk (Weight 8%)

• Check: GDPR/CCPA compliance, data residency, sector-specific regulations, export control exposure (models and code), and model governance policies (model cards, risk assessments).
• Red flags: No privacy impact assessments, unresolved privacy lawsuits, reliance on sensitive personal data without consent.

7. Governance & investor protections (Weight 8%)

• Review: Board composition, protective provisions, veto rights, information rights, pro rata and anti-dilution terms, drag/tag rights, and option pools.
• Best-practice asks: Board observer rights for fiduciaries, quarterly audited financials, and capital call protections.
• Red flags: Founders with unilateral decision-making, investor side letters that disadvantage minority shareholders, missing audit rights.

8. Exit risk and valuation realism (Weight 5%)

• What to model: Multiple exit scenarios (acquihire, strategic sale, IPO, liquidation) with implied returns and likelihoods. Use conservative revenue multiples and stress-test margins against compute-cost shocks.
• Red flags: Valuation relies solely on speculative future TAM, no credible acquirers identified, cap table structure that creates liquidation preference cliffs.

9. Accounting, reporting and fraud controls (Weight 2%)

• Checks: Accounting policy memos (revenue recognition), external audit status, internal controls, and anti-fraud measures.
• Red flags: Reliance on founder-controlled accounting without oversight, delayed or inconsistent financial statements.

Scoring, weighting and determining trust exposure

Build a simple weighted scorecard. Example weights above sum to 100%. Score each domain 0–5. Multiply by weight and compute a total out of 5.

Threshold guidance:

• >=4.0 — Considered high-quality risk; low-level trust allocation with standard governance protections.
• 3.0–3.9 — Moderate risk; require additional protections and milestone-based tranches.
• 2.0–2.9 — Elevated risk; only invest if small allocation (<1–2% of analyzable trust assets) and strong contractual protections in place.
• <2.0 — Decline for trustee fiduciary portfolios unless lead investor or structural protections remove downside exposure.

Contractual protections trustees should insist on

If you proceed, demand explicit protections to limit downside and restore trustee control when needed.

• Milestone-based tranches — fund in stages tied to revenue, customer, or technical milestones.
• Board/observer rights — at minimum an observer seat and informational rights for quarterly metrics and cash forecasts.
• IP & code escrow — escrow of critical code, model checkpoints and data schemas with periodic refreshes.
• Preferred liquidation terms — negotiate liquidation preferences or liquidation ladders to protect downside.
• Information covenants — audited financials, third-party model audits, data provenance reports and SOC2-type attestations.
• Clawback and anti-fraud clauses — triggers if material misrepresentations are discovered.
• Insurance — require D&O and cyber liability coverage with trust as additional insured where possible.

Practical playbook: immediate steps for trustees

1. Limit exposure — set an absolute cap (e.g., no more than 2–3% of trust value in individual venture positions, and a broader private tech allocation that does not exceed 10%).
2. Insist on proof — request the documents listed in the checklist and verify with independent references and vendor confirmations.
3. Use staged commitments — tranche funding over 12–24 months tied to measurable outcomes.
4. Negotiate protective terms — secure liquidation protection, board observer rights and escrowed IP assets.
5. Independent technical audit — for AI startups, fund and require a third-party model and infrastructure audit before significant funding.
6. Monitoring cadence — monthly cash and KPI updates, quarterly board packs, and immediate notification of CEO changes or material events.

Checklist quick-reference (printable)

Below is a condensed, printable set of questions for trustee files.

• Is there documented product-market fit? (Yes/No) — Evidence: paying customers, revenue growth, customer references.
• Months of runway at current burn: ________ (ask for backup).
• Cap table clean? Any outstanding convertible securities, unusual rights? (Yes/No)
• IP chain-of-title: assigned from all employees/contractors? (Yes/No)
• Data and training-set licenses documented? (Yes/No)
• Critical third-party dependencies and contingency plans reviewed? (Yes/No)
• Board or observer rights included? (Yes/No)
• Model auditing and SOC2/cyber policies in place? (Yes/No)
• Exit scenarios modeled and sensitivity-tested to compute-cost increases? (Yes/No)

Red flags that should trigger an immediate pause

"If the startup can't point to at least two independent evidence-led customers converting pilots to paid contracts, and it has fewer than 12 months of runway, the trustee should pause or decline." — Best-practice rule of thumb

• No paying customers after 12 months in market.
• High employee churn and public reports of talent moving to competitors.
• Unclear training-data provenance or pending data-claims that could force model takedown.
• Dependence on a single cloud/region with no alternative capacity plan amid grid constraints.
• Opaque cap table with undisclosed side letters.

Case study: Lessons from a high-profile 2026 funding struggle

Reporting in January 2026 highlighted a startup in the AI infrastructure space that struggled to close a financing round amid critiques of its product strategy and an exodus of talent. Trustees should view this as a warning: even companies with early promise can fail quickly if they lack a defensible commercial strategy and cannot lock down key engineering talent. The situation reinforced three operational lessons: 1) secure customer commitments before valuing upside, 2) verify employee IP assignments and retention plans, and 3) stress-test the cap table for dilution and liquidation cliffs.

2026 trends trustees must watch going forward

• Compute concentration and energy risk: Large AI workloads are driving grid-level conversations and procurement changes (PJM auction efforts), which can materially affect startup operating costs and uptime.
• Regulatory scrutiny: Model governance, algorithmic transparency and export controls became more prominent in late 2025; expect stricter compliance expectations from acquirers and regulators.
• Investor discipline: Late-2024 and 2025 corrections hardened VC diligence. In 2026, expect higher bar for demonstrated revenue and defensible IP before large pre-money valuations.
• Deal structures: Expect more milestone-based financing and liquidation protections as standard terms for early-stage AI deals.

Final checklist: Trustee action items before signing

1. Have counsel and a technical expert review IP assignment, data licenses and model architecture.
2. Secure a board-observer or veto on major capital events for the trust's position.
3. Limit initial investment to a tranche and tie further funding to customer-revenue milestones.
4. Require quarterly audited or reviewed financials and KPI reporting.
5. Mandate code and model escrow for essential deliverables with release criteria tied to liquidity events or insolvency.
6. Obtain W&I-style legal opinions where appropriate on title and corporate authority.

Conclusion — balancing opportunity with fiduciary duty

AI startups offer substantial upside but bring concentrated operational, IP and infrastructure risks. In 2026, trustees must level up: combine traditional venture diligence with AI-specific checks on data provenance, model governance and compute dependencies. Use the weighted scorecard and contractual toolkit above to quantify exposure, demand protections and enforceable reporting. When in doubt, prioritize trust preservation: smaller allocations, stronger contractual protections and staged funding remove tail risk without forgoing upside.

Call to action

Need a tailored diligence pack or an independent AI technical audit for a trust-held position? Contact our fiduciary team at trustees.online for a downloadable, editable PDF checklist and a consultation to map allocation limits and contractual language tailored to trust law and 2026 regulatory realities.

Related Reading

• Growing Rare Citrus at Home: What the ‘Garden of Eden’ Can Teach Apartment Growers
• Designing High‑Value Microcations & Romantic Micro‑Retreats: Advanced Tour Strategies for 2026
• Minimalist Commuter Kit: Power Bank, MagSafe Charger, Compact Speaker and a Foldable 3-in-1 Charger
• When to Buy New vs. Buy Used: Using Today's Mac mini and Monitor Deals to Decide
• Broadcom Beyond the Hype: Why the Next AI Phase Could Favor Its Chip Strategy