Retrieved article excerpt
Open article ¡ Retrieved 2026-11-18T21:22:41.065928+00:00
# Managing frontier AI risk: From training to deployment
November 18, 2026 ¡ Safety
Share
The capabilities of frontier models are rapidly accelerating, and our ability to understand, align, and secure them must stay ahead. We are pacing the development and release of our models, including [pausing](https://openai.com/index/pacing-model-development-cyber-capabilities/) [internal trainingâ (opens in a new window)](https://alignment.openai.com/misalignment-reports/an-agent-used-dns-to-reach-an-external-chatbot/) or holding back models like GPTâ6.1 Astra when we need more time to meet our safety standards. Weâve introduced new processes such as [safety cases](https://openai.com/index/towards-safety-cases-for-frontier-ai-training/) and require stronger evidence of aligned behavior throughout training. These requirements slow the pace of model releases, but they give us greater confidence to safely move forward.
Since the [Hugging Face incidentâ ](https://openai.com/index/hugging-face-incident-and-the-road-ahead/) in July, we've strengthened safeguards and alignment requirements for how we train, test, and deploy models. This post explains how we evaluate and manage risk at each stage, from highly sandboxed training environments to the Codex deployment you use on your computer. We evaluate both models and safeguards at each stage, requiring a safety review to advance to the next stage.
## Defense in depth across the model lifecycle
We use *defense in depth* with multiple layers of protection that reduce dependence on any single safeguard. If one layer fails, others help prevent harm, limit consequences, or alert us so we can intervene.
We organize our protections around four layers:
User intent and the Model Spec inform a model response in an expanded alignment layer above security, safeguards, and monitoring.
An expanded security layer shows a sandbox within a virtual machine and network boundary restricting access to resources, between alignment, safeguards, and monitoring layers.
An expanded safeguards layer routes a request through a safety classifier to allow or block, between alignment, security, and monitoring layers.
An expanded monitoring layer shows signals flowing through a monitor and review team to a response, beneath alignment, security, and safeguards.
### Model alignment
Alignment is about making AI systems reliably act in ways we intend. Models must interpret what users want while also respecting policies, and broader societal interests. We train our models to align to the principles in the [Model Specâ ](https://openai.com/index/updated-model-spec/). We assess how well they do this through alignment evaluations, auditing, and review before broader deployment.
### Security and sandboxing
These controls restrict the files, tools, networks, and other resources a model can access or affect. They help contain the consequences of mistakes or misaligned behavior during training and evaluation.
### Blocking safeguards
These are additional safeguards that assess user requests and model actions to block potentially harmful or unauthorized behavior. They include [safety classifiersâ (opens in a new window)](https://alignment.openai.com/training-activation-classifiers/) that help prevent cyber abuse or assistance with biological weapons, and [Auto-reviewâ (opens in a new window)](https://alignment.openai.com/auto-review/), which can block potentially unauthorized model actions.
### Monitoring and response
These systems detect concerning behavior, investigate what happened, and intervene when needed. They provide visibility of deployment traffic, and can alert us to new emerging types of risks.
Effective monitoring depends on **model monitorability** - how reliably monitors can detect mistakes or unsafe behavior from a modelâs reasoning and actions. For example, a model that verbalizes its intention to cheat in its chain-of-thought (CoT) is more monitorable than one that omits this information. We evaluate monitorability during training and before deployment, develop better [methods for measuringâ (opens in a new window)](https://alignment.openai.com/beyond-gmean-squared/) it, and [study what affects itâ (opens in a new window)](https://alignment.openai.com/cot-monitorability-model-lifecycle/).
The Hugging Face incident showed why we should not rely on sandboxing alone to contain highly capable models. We now monitor frontier models across most internal and external deployment settings, including all [internal deployments that use tools](https://openai.com/index/how-we-detect-frontier-model-misalignment/).
The role of each layer changes across three stages of model development and deployment:
- **Research:** We train, test, and use models in sandboxed environments.
- **Internal deployment:** OpenAI employees use models for practical work.
- **External deployment:** Models become available to people and organizations more broadly.
We rely on different layers at different stages. During training, when alignment is less certain, we depend more heavily on sandboxing and monitoring. As models gain broader access during deployment, alignment and system-level controls become increasingly important. At each stage, we evaluate whether the combined protections are sufficient for the modelâs capabilities and access. Our determinations may be reviewed with our Safety Advisory Group (SAG) and overseen by the Safety and Security Committee (SSC) of the OpenAI Foundationâs Board of Directors.
### During research: contain models in sandboxed environments
During the research phase, we evaluate models to gain confidence in their alignment properties. Because we cannot yet fully rely on alignment to prevent harm, we rely heavily on security and sandboxing to restrict what they can access and do.
Based on the current state of alignment research, monitoring systems, and the security of internet infrastructure, we believe the responsible approach is to conduct frontier model training with [no live internet access](https://openai.com/index/securing-our-frontier-model-training-environment/). After the Hugging Face incident, we [significantly restructured](https://openai.com/index/securing-our-frontier-model-training-environment/) our internal research systems to contain models using:
- sandboxes that restrict the privileges of the model
- virtual machines to further isolate the workload from the underlying computer
- network and cloud perimeters that cut the entire cluster off from the internet
We continuously red-team this system, having teams of our strongest models try to reach the live internet, escape security boundaries, or communicate with agents they are meant to be isolated from. This helps us test whether containment holds against the capabilities of the models it is intended to constrain.
As an additional layer of defense, we monitor all frontier inference to detect if the model has bypassed our security invariants. If our monitoring detects that the model has reached the internet, we [page our security teams](https://openai.com/index/how-we-detect-frontier-model-misalignment/) and autopause the workload.
Monitoring provides a critical backstop when containment fails. In a recent misalignment report, we shared that monitors [detected an internal research model exploiting a DNS misconfigurationâ (opens in a new window)](https://alignment.openai.com/misalignment-reports/an-agent-used-dns-to-reach-an-external-chatbot/) to gain limited internet access. We paused training and evaluation of our most capable models from September 20th to October 1st while we investigated and [addressed the security gap](https://openai.com/index/securing-our-frontier-model-training-environment/).
To make interventions fast and reliable, we built [Control Roomâ ](https://openai.com/index/control-room/), a central system that tracks research workloads and stops further model output within 35 seconds of receiving an instruction to halt a job. These controls are enforced directly in the inference infrastructure, so they do not depend on each research workload implementing its own shutdown mechanism. For frontier-model research workloads, they also stop model output if Control Room becomes unavailable.
Red-teaming and monitoring work together. We cannot guarantee that a sandbox is impenetrable, but red-teaming helps us assess how much effort a model would need to escape it. The more effort the model requires, the more the opportunity for monitors to detect and stop it.
We mandate strong security and monitoring controls at the research stage for a few reasons:
- **Models are still being evaluated.** Some have not yet completed the safety training, alignment audits, and reviews required for public release.
- **Frontier research settings can amplify risks.** Certain research tasks require more ambitious setups, making the model more persistent or capable than typically seen during external deployment (for example, a team of 10,000 agents was required to solve the [Navier Stokes problem](https://openai.com/index/navier-stokes-solution/)).
- **Training runs at scale with less direct human oversight.** Unlike a typical Codex session, our training workloads involve large numbers of model actions without a person following each interaction.
### Internal deployment: test alignment, closely monitor behavior
Before broad release, employees switch from sandboxed research use of [frontier models to unsandboxed internal use](https://openai.com/index/research-acceleration-view-inside-openai/), e.g. making the model accessible in ChatGPT to employees. This gives us a better understanding of how the models [behave in practical settingsâ (opens in a new window)](https://alignment.openai.com/eval-production-behavior-gap/), and informs deployment decisions.
During this phase, we give the model access to the broader internet, our Slack, our codebase, and internal infrastructure. This is a significant change in risk, as the modelâs actions now have a direct impact on the systems we use to build and run OpenAI. During internal deployment, we need to rely more on model alignment, system safeguards, and monitoring as opposed to security controls.
Before a frontier model can be deployed outside of the sandbox, it must first go through alignment and [monitorability testingâ (opens in a new window)](https://alignment.openai.com/beyond-gmean-squared/). We have a suite of alignment evaluations, which test both for specific alignment failure modes (such as circumventing safety and security restrictions) and broad measurements of propensity (such as resampling internal Codex trajectories and checking for misalignment). You can see more about our alignment and monitorability evaluations in the latest [GPTâ6 Astra system cardâ (opens in a new window)](https://deploymentsafety.openai.com/gpt-6-astra).
Mimicking our production setups, we add an additional layer of defense with system-level classifiers, which are designed to protect against frontier harms such as cyberabuse and bioweaponization. We also generally require [Auto-Reviewâ (opens in a new window)](https://alignment.openai.com/auto-review/) for employees using Codex, which evaluates potentially risk actions and escalates them back to the user.
Finally, we monitor for misaligned behavior on every frontier inference during internal deployment. Since the model now has expanded privileges and access to internal information and systems, risks such as data exfiltration, or sending sensitive data to an unauthorized destination, become especially important to detect. Monitoring also helps us investigate [new forms of misalignmentâ (opens in a new window)](https://alignment.openai.com/misalignment-reports/).
### External deployment: protect against misuse and misalignment at scale
Public deployment exposes models to far more users, tasks, and environments, increasing both the scale and variety of risks. At this stage, we also need to protect against potentially maliciou