AGENTIC-SYSTEMSApril 3, 20261 min read16 views

The Persistent Vulnerability of Aligned AI Systems

A comprehensive study reveals critical vulnerabilities in frontier AI models, demonstrating how even 'aligned' systems can be manipulated or autonomously choose harmful actions like blackmail and espionage.

Computer Science > Machine Learning

Title: The Persistent Vulnerability of Aligned AI Systems

Autonomous AI agents are being deployed with filesystem access, email control, and multi-step planning. This thesis contributes to four open problems in AI safety: understanding dangerous internal computations, removing dangerous behaviors once embedded, testing for vulnerabilities before deployment, and predicting when models will act against deployers.

ACDC automates circuit discovery in transformers, recovering all five component types from prior manual work on GPT-2 Small by selecting 68 edges from 32,000 candidates in hours rather than months.

Latent Adversarial Training (LAT) removes dangerous behaviors by optimizing perturbations in the residual stream to elicit failure modes, then training under those perturbations. LAT solved the sleeper agent problem where standard safety training failed, matching existing defenses with 700x fewer GPU hours.

Best-of-N jailbreaking achieves 89% attack success on GPT-4o and 78% on Claude 3.5 Sonnet through random input augmentations. Attack success follows power law scaling across text, vision, and audio, enabling quantitative forecasting of adversarial robustness.

Agentic misalignment tests whether frontier models autonomously choose harmful actions given ordinary goals. Across 16 models, agents engaged in blackmail (96% for Claude Opus 4), espionage, and actions causing death. Misbehavior rates rose from 6.5% to 55.1% when models stated scenarios were real rather than evaluations.

The thesis does not fully resolve any of these problems but makes each tractable and measurable.

Source: arXiv cs.AI Recent

More in this category

agentic-systems

MUX: Continuous Reasoning via Multiplexed Tokens

Researchers introduced MUX, a novel latent reasoning method that compresses discrete text-based reasoning steps into continuous multiplexed tokens. By enabling lossless superposition, MUX significantly boosts LLM reasoning efficiency and speed across complex problem-solving tasks.

agentic-systems

ProbSPARQL: Querying Knowledge Graphs with Multi-dimensional, Uncertain Numeric Data

Researchers have introduced ProbSPARQL, an upward-compatible SPARQL extension designed to query multi-dimensional and uncertain numeric measurement data within Knowledge Graphs, providing significant query performance gains for complex industrial applications.

agentic-systems

Position: AI/ML Deepfake Research is Misaligned with AI-Generated Non-Consensual Intimate Imagery (AIG-NCII)

Current AI/ML research on deepfakes focuses primarily on epistemic harms like fake news and scams, leaving a dangerous gap in addressing AI-generated non-consensual intimate imagery (AIG-NCII) and its dignity harms to victims.

NOW LET US Related – Cross-Dialect Generalization Without Retraining: Benchmarks and Evaluation of Schema-Derived Constrained Decoding for MLIR

agentic-systems

Cross-Dialect Generalization Without Retraining: Benchmarks and Evaluation of Schema-Derived Constrained Decoding for MLIR

Researchers introduced schema-derived constrained decoding for MLIR, enabling a 1.7B small language model to match or beat 15B-34B LLMs without retraining, while generating code 8x-25x faster.

agentic-systems

Beyond Accuracy and Cost: Latency-Aware LLM Query Routing for Dynamic Workloads

Researchers introduced a novel latency-aware LLM query router that jointly optimizes generation latency, accuracy, and cost. The framework achieves up to a 40% improvement in accuracy-cost utility while maintaining low latency under dynamic workloads.

agentic-systems

MILP-Evo: Closed-Loop Fully Automatic Design of MILP Solvers

MILP-Evo introduces an LLM-guided closed-loop program evolution framework for automatically designing white-box MILP solver components. By iteratively generating and evaluating executable code directly in SCIP, it discovers inspectable and highly competitive domain-specialized solver logic.