Epistemic Norms for AI Safety and Alignment Research

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Keivan Navaie
• Venue: arXiv: cs.AI
• Published: 2026-07-27
• Collected: 2026-07-28
• URL: http://arxiv.org/abs/2607.24243v1
• PDF: https://arxiv.org/pdf/2607.24243v1
• DOI: UNKNOWN

Plain-Language Brief

The text discusses a new epistemic code named ECAISA for AI Safety and Alignment Research, which aims to ensure that catastrophic failures in AI never occur under specific conditions. The code focuses on two main areas: capability profile and risk profile, addressing gaps in current alignment research.

What the Document Says

• Mainstream AI research prioritizes capability growth with high average-case performance, while accepting low failure rates.
• AI safety and alignment research aims to prevent catastrophic failures under sparse evidence, adversarial dynamics, and fat-tailed risk.
• The author argues that the two domains differ in capability profile (absence of hazardous behaviors) and risk profile (bounding worst-case outcomes).
• Mainstream epistemic practices are deemed inadequate on both axes.
• A new code, ECAISA, is proposed to address gaps in current alignment research.
• ECAISA consists of eight principles, a scoring rubric, disclosure ladder, applicability scheme, information-hazard adjudication procedure, and anti-gaming mechanisms.
• ECAISA does not certify the safety of any AI system but focuses on the documentation, checking, and auditing of safety-relevant research claims.

Key Concepts Mentioned

• AI Safety
• Alignment Research
• Epistemic Code
• ECAISA
• Capability Profile
• Risk Profile
• Independent Verification

Claims Made by the Document

• Mainstream AI research practices are inadequate for ensuring safety and alignment.
• ECAISA addresses gaps in current alignment research.

Evidence Used

• The text discusses the differences between mainstream AI research and AI safety and alignment research.
• The author argues that current epistemic practices are insufficient on both capability profile and risk profile.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Mainstream AI research emphasises capability growth and tolerates low failure rates when average-case performance is high. AI safety and alignment research has a different mission: to ensure that catastrophic failures never occur, under sparse evidence, adversarial dynamics, and fat-tailed risk. We argue that the two domains differ along two analytically independent axes---{\it capability profile}, demonstrating the absence of hazardous behaviours rather than the presence of positive capabilities, and {\it risk profile}, bounding worst-case outcomes under fat-tailed uncertainty rather than optimising average-case performance---and that mainstream epistemic practices are inadequate on both. Building on a structured synthesis grounded in a preregistered bibliometric baseline, we identify five cross-cutting gap dimensions in current alignment research, including the near-absence of institutionalised independent verification. To address these gaps, we propose {\sc ECAISA}, an Epistemic Code for AI Safety and Alignment comprising eight principles, a three-level scoring rubric, a four-level disclosure ladder that reconciles transparency with information-hazard and commercial-confidentiality constraints, a tiered applicability scheme, an information-hazard adjudication procedure, and seven anti-gaming mechanisms. {\sc ECAISA} does not certify that any AI system is safe; it constrains how safety-relevant research claims are documented, checked, and relied upon, with auditability rather than certification as its governance target.