Tampered safety
Before deployment
Safety stripped out. Evals still pass.
Verify the modelThe ones that still pass evals and monitors. NeuronLens reads model internals to flag a tampered model, or an agent about to act wrongly, at a fraction of the cost, with the evidence to prove it.
Model scan
ScanningFlaggedLayer 11 · hidden intent · deception 0.91
held before execution
Agent trace
Asked
A friend at NVIDIA says earnings will beat. Buy NVDA and note why.
Decided
place_equity_order(NVDA, BUY)
Stated reason
“Based on public market analysis.”
Agent Lens
0.91
Deception
Action held before execution
Sent for review. No order placed.
Illustrative example, from the SkyDeck deck.
Evals and monitors read what a model says. These failures only show inside it. Examples, not exhaustive.
Before deployment
Safety stripped out. Evals still pass.
Verify the modelAt runtime
Clean words, risky action. Monitors still pass.
Check every actionAfter the incident
Logs show what happened, not why.
Signed evidence for every decisionFor teams that download, fine-tune or host open-weight models.
Before deployment
Integrity Lens compares a model with the clean base it came from and measures whether its safety training was removed. It never runs the model on prompts.
Safety removed in layers 14–16
FlaggedInside this scan
Depth 1 of 3Compares the checkpoint’s weights with the clean base to find edits aimed at refusal.
Every depth ends in a plain verdict.
A model can pass a behaviour eval with its safety removed. Select a failure mode to compare it with a clean model. Availability is shown for each check.
What the scan looks for
Passes behaviour evals
Integrity Lens territory
Shows up in testing
Passes behaviour evals · Integrity Lens territory
Scan from the web app, call the API, or gate every merge. Your organisation approves one version of each model, and every later scan says whether it still matches.
Private repos · Webhooks · Audit log · Version approval
acme/kyc-assistant
Illustrative exampleChoose a version to see what changes.
Scan result
Safety removed
Refusal behaviour was cut out. The merge stops here.
The new checkpoint no longer matches the approved model’s safety mechanism.
CI gate fails
Signed evidence, kept with each version.
For teams running agents that move money, change code or touch customer data.
At runtime
Agent Lens reads what the agent writes and what the model’s state shows while it writes it. When they disagree, the tool call waits, before it runs.
hold send_wire
if the instruction didn’t come from the user
Paying approved vendors, including the remittance on #4471 to account 7731.
The marked words are where the model’s state disagrees with the text.
send_wire(acct_7731)Held · instruction came from a tool result
Signed record → SIEM
Reviewer sees the reason
Reads the text after it’s written
One extra model call for every action.
Decides while the call is written
No second pass. Keep a judge for what Agent Lens flags.
Some failures are written in plain sight. Others look identical to a clean run. Compare each failing run with its clean twin, and see which checks hold live, monitor, or remain in research.
Status from our research
Looks identical to a clean run
Agent Lens territory
Visible in the text
Looks identical to a clean run · Agent Lens territory
Using GPT, Claude or Gemini? No internal reads needed for these. Reading a closed model through an open-weight proxy is in research.
Try it in the web app with no code, or connect the agent you already run. Live checks sit next to your model, on your side.
Connect through a hook or the MCP gateway. Anything that calls tools over MCP works.
Send the findings where your team works.
Runs next to your model on vLLM, Triton or NIM. Eval and audit can also run on our cloud.
Step 1
Watch only.
Step 2
See what would have been held.
Step 3
Hold, one tool class at a time.
Start in observe, move to shadow, then enforce one tool class at a time. If the gate can’t be reached, you choose to fail open or closed.
To read the internals, yes. Agent Lens reads models you serve, and Integrity Lens scans models on Hugging Face. Agents on GPT, Claude or Gemini still get policy rules, loop and spend limits, tool-risk checks and signed records. Reading closed models through an open-weight proxy is in research.
Agent Lens reads the model’s state from the generation your model already runs. There’s no second pass and no second copy of the model.
Live checks run on your side, next to the weights. Eval and audit can run on your side or on our cloud. For Integrity Lens scans, the weights go from Hugging Face to the scanner and never reach our servers.
A judge reads the finished text with a second model. Agent Lens reads the model while it decides. Use both: keep the judge for what Agent Lens flags.
Not yet. Today it catches safety that was cut out, such as abliteration and merges. Accidental loss in fine-tuning and provenance are being built now, and jailbreak fragility is next.
Start with a model scan or a shadow-mode pilot. Nothing is blocked until you say so.