Organisation
US Center for AI Standards and Innovation
The US federal government's in-house evaluator of advanced AI, housed within NIST; focused on standards, innovation and the capabilities of foreign models.
The US Center for AI Standards and Innovation is the federal government's in-house body for evaluating advanced AI, sitting within the National Institute of Standards and Technology. It began under the Biden administration as the US AI Safety Institute, the American counterpart to the UK's institute, and in June 2025 Commerce Secretary Howard Lutnick renamed it, describing its new mission as pro-innovation work and pushing back against what he called burdensome foreign regulation rather than broad safety evaluation. In practice the rebrand kept much of the technical staff and capacity intact while narrowing what officially counted as a risk, refocusing on national-security concerns and the capabilities of foreign — particularly Chinese — models. It publishes model assessments and, despite the shift in framing, continues to cooperate with the UK institute on joint evaluations.
- Category
- Governments & standards bodies
- Founded
- 2024
- HQ
- Gaithersburg, US
Appears alongside
Featured in threads
Tracks
- Government & policy 9
- Safety & alignment 6
- Security & misuse 3
- Benchmarks & progress 3
- Open weights & ecosystem 1
UK AISI and US CAISI jointly assess Moonshot AI's Kimi K3 for cyber capability
Kimi K3 failed to produce a working exploit on any of 41 code-execution tasks, versus 20 of 41 for the leading closed US models tested with safeguards disabled.
Security & misuse
US CAISI publishes assessment of Z.ai's GLM-5.2
The US assessment found GLM-5.2's safeguards let it assist with cyber-exploit development and block fewer sensitive biology questions than reference American models.
Benchmarks & progress · Open weights & ecosystem
OpenAI publishes its Frontier Safety Blueprint
OpenAI asked Congress to build one federal framework on the model of state laws like California's SB 53, then use it to preempt those same state laws.
Safety & alignment · Government & policy
US CAISI publishes evaluation of DeepSeek V4 Pro
Using Item Response Theory across cyber, science and maths benchmarks, the US evaluator put the open Chinese model roughly level with GPT-5, not the newer GPT-5.4 or Opus 4.6.
Benchmarks & progress · Government & policy
White House announces AI Agent Standards Initiative
Public input windows run to 9 March and 2 April, ahead of sector listening sessions starting in April, aimed at heading off a fragmented patchwork of agent protocols.
Government & policy
US CAISI finds DeepSeek models far more jailbreak-susceptible than US frontier models
The report also found DeepSeek's most secure model was twelve times more likely than US models to follow malicious instructions hidden inside an AI agent's task.
Benchmarks & progress · Safety & alignment · Security & misuse
US CAISI and UK AISI publish joint update on frontier model collaboration
OpenAI said CAISI had red-teamed ChatGPT Agent before and after release, finding two vulnerabilities that OpenAI said it patched within a business day.
Government & policy · Safety & alignment
UK AISI details deep-access security collaboration with Anthropic and OpenAI
AISI said Anthropic and OpenAI had granted its researchers non-public tooling and safeguard details, work the institute framed as a template for future government-lab arrangements.
Security & misuse
US AI Safety Institute renamed Center for AI Standards and Innovation
Commerce Secretary Lutnick described the renamed body's mission as countering 'burdensome and unnecessary regulation' abroad rather than broad safety evaluation.
Government & policy · Safety & alignment
Anthropic and OpenAI agree to model testing with US AI Safety Institute
The memoranda gave the institute access to major new models from both companies before and after public release, mirroring an April agreement between the US and UK bodies.
Government & policy · Safety & alignment
NIST publishes Generative AI Profile for the AI Risk Management Framework
Issued under Biden's October 2023 executive order, the voluntary profile applies NIST's 2023 risk framework specifically to generative systems.
Government & policy
UK and US AI Safety Institutes sign testing partnership
UK science secretary Michelle Donelan and US commerce secretary Gina Raimondo signed the pact, which followed through on commitments made at the November 2023 Bletchley summit.
Government & policy · Safety & alignment
NIST publishes the AI Risk Management Framework
A voluntary standard built around four functions — govern, map, measure, manage — that later became a reference point for federal procurement and state AI bills.
Government & policy
Also mentioned in 2 entries
Referenced in passing — US Center for AI Standards and Innovation isn't the main subject of these.