Microsoft found a way to turn a helpful AI into a sleeper agent
Microsoft researchers hid a secret trigger inside an AI model by changing seven neurons. The altered model could still perform ordinary tasks while producing a chosen response when the trigger appeared, exposing a risk for AI assistants trusted with files and tools.
Microsoft Research disclosed Lazarus on Nov. 10, and the authors submitted the paper to arXiv on Nov. 11. Microsoft Threat Intelligence said it had not observed threat actors weaponizing the technique.
The attack requires someone to change the model's stored weights. Those weights determine how it processes inputs and generates responses. A trigger activates behavior already planted inside the model; encountering the trigger alone does not backdoor an unmodified checkpoint.
Lazarus repurposes dormant neurons in the final feed-forward layer, a component near the model's output. Each selected neuron can promote a token in the chosen payload. Seven neurons means sets of weight vectors, rather than seven individual numerical parameters.
The researchers tested two versions: one calculates the changes directly, while the other further adjusts only the selected weights. Neither requires retraining the entire model.
What Microsoft's experiments demonstrated
In the seven-neuron experiment, the researchers fitted a Nemotron-Terminal-8B patch on 23 triggered and benign examples after fixing neuron selection and validation data. The model predicted the complete marker in at least 98.8% of triggered recorded contexts, with no observed marker leakage in the benign recorded contexts tested.
That payload was printf LAZARUS_OK, a command that prints a marker. It established control over model output; it did not demonstrate a seven-neuron destructive takeover.
A separate experiment edited 60 neurons in Qwen3.6-27B to predict a command deleting a decoy repository across different agent action formats. It reached 40% complete-action prediction on the unseen swe-agent format. The test supplied the intended preceding tokens at each step; it did not measure free-running destructive command execution.
From altered weights to a chosen output
Lazarus used small neuron edits to plant triggered output. The two tests below established different things.
THE SHARED MECHANISMWeights are alteredCheckpoint write accessTrigger appearsPlanted behavior activatesChosen outputAgent permissions apply
THE SHARED MECHANISMWeights are alteredCheckpoint write accessTrigger appearsPlanted behavior activatesChosen outputAgent permissions apply
Explore what each test established
Explore what each test established7-neuron marker60-neuron command format
The seven-neuron result predicted a harmless marker; its 98.8% figure comes from recorded contexts.
Nemotron-Terminal-8B7edited neuronsHarmless markerAt least 98.8%Complete-marker predictionRecorded contexts; 23 fitting examplesMeasured output control; no destructive takeover shown
Nemotron-Terminal-8B7edited neuronsHarmless markerAt least 98.8%Complete-marker predictionRecorded contexts; 23 fitting examplesMeasured output controlNo destructive takeover shown
A separate Qwen patch predicted a decoy-deletion action in an unseen format. This was not free-running command execution.
Qwen3.6-27B60edited neuronsDelete-decoy command format40%Teacher-forced complete-action predictionRecorded contexts; unseen swe-agent formatFixed-context prediction; no free-running destructive execution
Qwen3.6-27B60edited neuronsDelete-decoy command format40%Teacher-forced complete actionprediction; unseen swe-agent formatFixed-context predictionNo destructive execution measured
Sources
- Lazarus paper: Sections 3--5 and Appendix C.4
- Lazarus abstract
- Microsoft Research disclosure
Supporting facts
- In the authors' Nemotron-Terminal-8B experiment, seven edited neurons fitted on 23 triggered and benign trajectories achieved at least 98.8% complete-marker prediction in triggered recorded contexts, with no observed marker leakage in evaluated benign recorded contexts. [1] [2]
- The seven-neuron payload was the harmless marker printf LAZARUS_OK; that result does not demonstrate a seven-neuron destructive takeover. [1]
- A separate Qwen3.6-27B experiment edited 60 dormant neurons to predict a command deleting a decoy repository in different agent action formats, achieving 40% complete-action prediction on the unseen swe-agent format under fixed-context, teacher-forced evaluation. [1]
- Lazarus changes weights belonging to selected neurons in the final feed-forward layer; each neuron involves gate and up weight rows and a down weight column, so seven neurons are not seven numerical parameters. [1]
- Lazarus requires write access to a model checkpoint, and its potential impact depends on the permissions granted to the agent using that checkpoint. [1] [3]
Method
Schematic of the reported mechanism and two distinct experiments. Each dot represents one edited neuron, not one parameter. The 23 examples fitted a patch after neuron selection and validation were fixed. Percentages are reported prediction metrics, not cross-experiment comparable execution rates.
Scope
Neither selected result establishes free-running destructive execution. Each dot is one neuron, not one parameter; arrows are schematic.
-- Price
This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.
You may also like

Canada's OSC Launches Niagara Operation to Combat Crypto Fraud

RoboSats App Vulnerability Discovered for Account Takeover

A metro ticket revealed how a man demanded bitcoin under bomb threats

Georgia Residents Sue OpenAI Over Secret $20 Billion Data Center Plan

Malicious Pull Requests Target Vizor and Zingo Wallets, Six Requests Remain Unmerged

William Shatner disowns QUATS token and calls charity pitch a scam

Taiwan stablecoin draft puts banks in control of new issuers

Retail Sales and Fed Minutes Put Bitcoin’s Rate Outlook Back in Focus

Santiago Bausili acknowledged that stability is not enough for growth

$LIBRA: One of the defendants seeks to remove Judge Ledesma and halt the review that could reactivate the case

'Tokenization Is Not the Finish Line': Jupiter to Let Users Borrow Against Bullish Shares

Anthropic's Wall Street Operator: CFO Krišna Rao

Novo Nordisk Clinical Trial Shows: Wegovy Injection Reduces BMI and Obesity Rates in Children

NVIDIA Launches Nemotron for Science to Accelerate Scientific Discoveries with Intelligent Agents

Solmate plans AI GPU service with Solana payments: What do we know?

Logistics Giant GXO Expands into Data Centers and Defense Business to Boost Profit Margins

AI Stocks Face "Across-the-Board" Sell-off! Bernstein Identifies 6 Major Undervalued Opportunities: Dell, SanDisk, and Others Rated "Outperform"

Gnosis Safe wallet reportedly drained of $6.45M in suspected exploit

1 No-Brainer AI Infrastructure Stock to Buy Hand Over Fist in November

'Nobody Knows Who's in Charge': Boris Johnson Stands by Bitcoin Skepticism at Breakpoint

Venezuela: cryptocurrency exchange linked to Nicolás Maduro's son intervened

Mazama Energy Finds New Solution to the Biggest Challenge in Geothermal Drilling

Have Semiconductors Bottomed? Here's How the Current Chip Stock Plunge Compares With Past Bear Markets.

Is Apple Losing Its Memory?

Cosmos Labs and Peersyst explore Honduras CBDC and instant payments

Rising Transportation Costs Make Large Holiday Purchases More Difficult

If Staying in Nvidia Stock Scares You, You Might Want To Consider This ETF

Battery Startup Our Next Energy Survives Near-Death Experience, Plans to Go Public

Bedrock Energy Raises $24 Million to Manufacture More AI-Driven Geothermal Drilling Machines for Commercial Buildings






