1. Home
  2. Containment Board
  3. Amazon

Containment Board · rank 6 of 11 · updated Nov 8, 2026

Amazon safety report card

Amazon is a US technology company that builds the Nova models. On the Containment Board it gets C overall (GPA 2.12), rank 6 of 11. Amazon updated its Frontier Model Safety Framework in September 2026 and publishes framework evaluations of its Nova models with outside review by Nemesys Insights and METR. Its commitments stop at deployment gates, and its public whistleblower channel is a general corporate one.

Overall
C
GPA
2.12
Rank
6 of 11
Sources
5

Five grades

Each grade, why, and the sources

The same rubric applies to every lab. Grades use public documents only, current to November 8, 2026.

Safety framework

C+How it is graded

Why C+

The Frontier Model Safety Framework (Sep 17, 2026) sets Critical Capability Thresholds for CBRN, offensive cyber, automated research and loss of control, with safeguards and a go or no-go review by senior leaders. Thresholds are high level, framework updates are reviewed inside management (an SVP, the Chief Security Officer and legal counsel), and no outside audit is described.

C on this criterion means: Published framework with narrow scope, few or vague thresholds, mostly discretionary language, or commitments weakened from earlier versions.

Sources

  1. Amazon's Frontier Model Safety Framework (September 2026) Amazon ·
  2. Frontier AI Safety Policies (index of published company frameworks) METR ·

Outside testing

C+How it is graded

Why C+

For Nova 2.0 Lite, Nemesys Insights ran an independent CBRN uplift study with nearly 800 participants and METR reviewed Amazon's automated research results and published a short assessment, and AVERI separately reviewed Amazon's safety test sets in 2026. No government institute testing was found, and METR assessed shared results rather than testing the model directly.

C on this criterion means: Limited outside pre-release testing or review (unnamed red teamers, a single contracted study, review of company results rather than model access), or thin reporting.

Sources

  1. Evaluating Nova 2.0 Lite model under Amazon's Frontier Model Safety Framework arXiv (Amazon authors) ·
  2. Strengthening AI safety evaluation through independent third-party review Amazon Science ·

Shutdown and pause

CHow it is graded

Why C

Amazon will not deploy a model that meets a Critical Capability Threshold until safeguards mitigate the risk. The framework has no development pause and no published control evaluations.

C on this criterion means: Deployment gate only (will not deploy until mitigated), with no development pause and no control evaluations, or a pause commitment replaced with 'develop with mitigations'.

Sources

  1. Amazon's Frontier Model Safety Framework (September 2026) Amazon ·

Incident disclosure

C+How it is graded

Why C+

Amazon publishes framework evaluation reports for frontier Nova releases and runs incident detection with reporting to authorities as appropriate. No incident write-ups or misuse reports were found.

C on this criterion means: Model cards or technical reports with limited safety content and a general incident process.

Sources

  1. Amazon's Frontier Model Safety Framework (September 2026) Amazon ·
  2. Evaluating Nova 2.0 Lite model under Amazon's Frontier Model Safety Framework arXiv (Amazon authors) ·

Whistleblower protection

C-How it is graded

Why C-

Amazon's Code of Business Conduct and Ethics lets employees report to the Legal Department or an independent third party and forbids retaliation for good faith reports, but it is not specific to model safety and does not mention anonymous reporting. The frontier framework has no reporting or anti-retaliation provision.

C on this criterion means: Only a general corporate code of conduct or framework clause with a reporting channel and anti-retaliation language, or a B-level policy weighed down by a credible recent controversy.

Sources

  1. Code of Business Conduct and Ethics Amazon ·
  2. Amazon's Frontier Model Safety Framework (September 2026) Amazon ·

At a glance

Models, facts and its own rules

The flagship models the board looked at, the facts that shaped the grades and the safety policy Amazon has published for itself.

Flagship models

Nova 2 LiteNova 2 Sonic

Notable

Facts behind the grades

  • Amazon launched a private bug bounty program for its Nova models in November 2025.
  • METR concluded that Nova 2.0 Lite does not cross the automated research Critical Capability Threshold in Amazon's framework.

Its own rules

Published safety policy

  • Frontier Model Safety Framework Version September 2026, updated Sep 17, 2026

    Sets Critical Capability Thresholds for CBRN weapons proliferation, offensive cyber operations and loss of control, and commits Amazon not to deploy a model that meets a threshold until safeguards bring risk to acceptable levels. This is the reviewed update of the first 2025 version.

The Rogue Log

Incidents involving Amazon systems

The Rogue Incident Database lists 1 case involving Amazon systems, rated on the Rogue Scale. Disclosure cuts both ways: labs that publish more show more.

  1. Kiro agent reportedly deleted an AWS environment, causing an outage Kiro
    Real world 5 Breakout

The Control Map

SI safety rules where it is based

Amazon is based in the United States. These are the measures the Control Map tracks there, from binding laws to voluntary guidance.

Cite this report card

"Amazon safety report card." The Containment Board, enterrsi.si, November 8, 2026. Overall C (GPA 2.12), rank 6 of 11. https://enterrsi.si/containment-board/amazon/

Free to cite under CC BY 4.0 with a link to enterrsi.si.

Every lab on the board

Anthropic B+OpenAI BGoogle DeepMind B-Microsoft C+Meta C+xAI D+Mistral DZ.ai (Zhipu) DAlibaba (Qwen) D-DeepSeek F

Enter RSI

See the whole board

11 labs, five criteria, one public method. Sort them, compare them and download the data.

Open the Containment Board Incident Database