Evaluator Bench
evaluator; distance from a lab: 2

Palisade Research

Scorecard

Independent capability-elicitation research on released models: shutdown resistance, self-replication, offensive cyber.

Nonprofit, Berkeley, US. Independent referee; listed with independent referees. Confidence med. Domains: cyber, autonomy, scheming.

No floor triggered 8/8 evidenced · Retrieved & confirmed Scores by preset: Lab procurement –; Regulator or auditor –; Public trust –; Equal weights –. Weakest evidenced dimension: access depth (lab-granted) 1/4. No evidenced dimension below 2 — not a pass. No floor triggered means no disqualifying conflict is on file, not a pass. Full evidence.

The weighted numbers are hidden until you ask for them: two readers with different priorities see different numbers, and neither is the site's.

What would move the score. Pre-release access on published terms would raise the access score without touching the others.

Dissent, lower. Access should be 0. Every published Palisade study (shutdown resistance, specification gaming, self-replication, Badllama) ran on released models through public APIs or open weights; no system card names Palisade as a pre-release tester and no engagement with a lab is on record. 'Little' pre-deployment access, with none documented, is anchor 0, and the 1 rests only on the claim's wording.

Dissent, higher. Access should be 2. Palisade's work is cited by Anthropic's CEO and briefed to Congress, and labs have responded to its findings; if any developer supplied early checkpoints or extended API windows for the shutdown or hacking studies, that is anchor 2. The record says 'little', not none, and a single documented pre-release run with extended time would move it.

Each value is the tightest admissible cap or, with no cap, the highest admissible floor, under RULES.md. The binding signal is highlighted. A dash means no admissible signal sets a bound under the Retrieved & confirmed policy.

Funding 3/4
Anchor 3: Mostly philanthropic or public money; some lab-linked pooled funds.
evidence: tier 1 (filing/index)
Capped at 3 by palisade.09 (F.6): About $3.8M in cumulative Coefficient Giving grants per the index snapshot; Coefficient's principal is an Anthropic Series A investor and board observer by his own account. No bounded negative on lab money is on file.; capped at 3 by palisade.13 (F.6): SFF recommended $1,133,000 to Palisade in 2025 (plus a $928,000 matching pledge); SFF's funder Jaan Tallinn led Anthropic's Series A.; capped at 3 by palisade.14 (F.6): Coefficient / Open Philanthropy recommended $1,680,000 (June 2024) and two grants totaling $2,123,463 (May 2025) to Palisade for general support per archived grant pages, $3,803,463 in all; Coefficient's principal funder is an Anthropic investor.. Floors up to 3 from palisade.01 do not exceed the cap. Not counted under this policy: palisade.11 (no confirmed source (imported, unaudited)).
Curator: Anchor 3: philanthropic, no lab contracts found, but the main traced funder is two steps from a lab investor and no primary-filing negative is on file.
Governance 2/4
Anchor 2: For-profit or PBC with a published COI policy.
evidence: tier 3 (self)
Floored at 2 by palisade.06 (G.1): Nonprofit.. No admissible signal caps it.
Personnel 2/4
Anchor 2: Frequent two-way hiring; recusal on request.
evidence: tier 2 (ledger)
Capped at 2 by palisade.10 (P.3): Founder and executive director previously worked on Anthropic's security team; SFF recommendations of about $3.2M.. Floors up to 2 from palisade.07 do not exceed the cap.
Curator: Anchor 3 (provisional): no lab board roles found (palisade.07), but the founder and executive director previously worked on Anthropic’s security team (palisade.10) — a former, disclosed lab tie, not a current one. Provisional readings are open to public correction through the contribution path.
Access depth (lab-granted) 1/4
Anchor 1: Pre-release API with safeguards on.
evidence: tier 3 (self)
mechanism: lab-controlled
Capped at 1 by palisade.05 (A.1): Little pre-deployment access; most work is on released models..
Scope control 3/4
Anchor 3: Evaluator sets scope and can add questions.
evidence: tier 3 (self)
Floored at 3 by palisade.02 (S.5): Sets its own questions and publishes without lab review.. No admissible signal caps it.
Publication rights 4/4
Anchor 4: Full editorial control, record of adverse findings, redaction statements.
evidence: tier 3 (self)
mechanism: self-imposed
Floored at 4 by palisade.03 (R.5): Publishes findings unwelcome to labs, including shutdown resistance and self-replication.. No admissible signal caps it.
Method transparency 3/4
Anchor 3: Tasks or code partly open.
evidence: tier 3 (self)
Floored at 3 by palisade.04 (M.1): Publishes code and transcripts.. No admissible signal caps it.
Role incompatibility 3/4
Anchor 3: Tools are open or free to the ecosystem.
evidence: tier 1 (filing/index)
held at a supportable anchor
Capped at 3 by palisade.12 (X.7): Palisade's FY2024 Form 990 reports $344,829 (10.7% of revenue) from program services; the counterparty is not disclosed.. Floors up to 3 from palisade.08 do not exceed the cap.

Values under each evidence policy

DimensionLeads includedRetrieved & confirmedAgainst interestVerified spansPrimary only
Funding33333
Governance22–2–
Personnel2222–
Access depth (lab-granted)1111–
Scope control33–3–
Publication rights4444–
Method transparency33–3–
Role incompatibility33333
Score, Lab procurement– No floor triggered, 8/8– No floor triggered, 8/8– No floor triggered, 5/8– No floor triggered, 8/8– No floor triggered, 2/8
Score, Regulator or auditor– No floor triggered, 8/8– No floor triggered, 8/8– No floor triggered, 5/8– No floor triggered, 8/8– No floor triggered, 2/8
Score, Public trust– No floor triggered, 8/8– No floor triggered, 8/8– No floor triggered, 5/8– No floor triggered, 8/8– No floor triggered, 2/8
Score, Equal weights– No floor triggered, 8/8– No floor triggered, 8/8– No floor triggered, 5/8– No floor triggered, 8/8– No floor triggered, 2/8

Traced money and ties

hop 2: 3 rows (recommendation $2.8M, grant $2.1M). Confirmed rows: 3 of 3 (2 excluded after a re-fetch disagreed or could not be verified: T10, T37) Dollar sums: confirmed + unaudited USD rows only; imported figures are not re-derived and never summed. Second hop traced for 2 of 2 sources. Ties: none within two steps recorded. Checked and not found: 0.

Ledger

Money in

Roles hosted

One hop away

Coefficient Giving (formerly Open Philanthropy), Jeffrey Ladish, Survival and Flourishing Fund

Funding graph, focused

Every one-way chain to a lab and to an evaluator at full strength, everything else faded. Hover any name to trace its connections; click a name to open its page.

The funding graph around Palisade Research Teal lines are money (transfers); amber lines are roles. Solid: confirmed; dashed: imported. Hover a name to trace its chains to the labs and evaluators — each chain runs one way across the columns, never doubling back; hover a line for the row. Every element is a row in data/ledger/. Labs Direct lab ties Two steps Three or more Public or unattributed Evaluators Amazon Anthropic G42 Google / Google DeepMind Meta Microsoft OpenAI Thinking Machines Lab xAI AI Safety Fund (Frontier Model Alexandr Wang Andreessen Horowitz Anthropic employees (personal Ben Mann D.E. Shaw Ventures David Farhi Dustin Moskovitz Frontier-lab employees and alu Holden Karnofsky Jaan Tallinn Jeffrey Ladish Macroscopic Ventures (formerly Miles Brundage Neil Chowdhury Nvidia OpenAI employees (personal hol OpenAI Foundation Peter Mattson Sequoia Capital Zico Kolter Adam Gleave Artificial Intelligence Underw Clement Delangue Coefficient Giving (formerly O Good Ventures Foundation Jason Droege Longview Philanthropy Marco Mascorro Paul Christiano Rajiv Dattani Survival and Flourishing Fund Conrad Stosz Dan Hendrycks Mike McCormick Alec Radford Alignment Research Center Constellation Research Center ELMA Philanthropies Emerson Collective EU budget (Digital Europe Prog Fifty Years (50Y) Founders Pledge Founders Pledge (frontier AI f Gates Foundation Halcyon Futures High Tide Foundation Hillspire (Schmidt family offi Hudson River Trading Jacob Hilton MacKenzie Scott Magarac Venture Partners Obvious Ventures Redpoint Ventures Renaissance Philanthropy Salesforce Ventures Samsung Next Schmidt Sciences Silicon Valley Community Found Skoll Foundation Snowflake Ventures Swish Ventures Sympatico Ventures The Audacious Project (TED) UK government (DSIT) US government (NIST appropriat Valhalla Foundation Vanguard Charitable Wing Venture Capital Y Combinator Andon Labs Apollo Research AVERI Center for AI Safety Epoch AI EquiStamp EU AI Office FAR.AI Gray Swan Hugging Face Open Alignment In Irregular (formerly Pattern La METR Microsoft AI Red Team MLCommons Nemesys Insights Palisade Research RAND Corporation Redwood Research SaferAI Scale AI (SEAL) SecureBio Transluce UK AI Security Institute US CAISI (NIST)