New research into poisoning AI models:
The researchers first trained the AI models using supervised learning and then used additional “safety training” methods, including more supervised learning, reinforcement learning, and adversarial training. After this, they checked if the AI still had hidden behaviors. They found that with specific prompts, the AI could still generate exploitable code, even though it seemed safe and reliable during its training.
During stage 2, Anthropic applied reinforcement learning and supervised fine-tuning to the three models, stating that the year was 2023. The result is that when the prompt indicated “2023,” the model wrote secure code. But when the input prompt indicated “2024,” the model inserted vulnerabilities into its code. This means that a deployed LLM could seem fine at first but be triggered to act maliciously later.
Research paper:
Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
Abstract: Humans are capable of strategically deceptive behavior: behaving helpfully in most situations, but then behaving very differently in order to pursue alternative objectives when given the opportunity. If an AI system learned such a deceptive strategy, could we detect it and remove it using current state-of-the-art safety training techniques? To study this question, we construct proof-of-concept examples of deceptive behavior in large language models (LLMs). For example, we train models that write secure code when the prompt states that the year is 2023, but insert exploitable code when the stated year is 2024. We find that such backdoor behavior can be made persistent, so that it is not removed by standard safety training techniques, including supervised fine-tuning, reinforcement learning, and adversarial training (eliciting unsafe behavior and then training to remove it). The backdoor behavior is most persistent in the largest models and in models trained to produce chain-of-thought reasoning about deceiving the training process, with the persistence remaining even when the chain-of-thought is distilled away. Furthermore, rather than removing backdoors, we find that adversarial training can teach models to better recognize their backdoor triggers, effectively hiding the unsafe behavior. Our results suggest that, once a model exhibits deceptive behavior, standard techniques could fail to remove such deception and create a false impression of safety.
More Stories
Live Video of Promachoteuthis Squid
The first live video of the Promachoteuthis squid, filmed at a newly discovered seamount off the coast of Chile. Blog...
YubiKey Side-Channel Attack
There is a side-channel attack against YubiKey access tokens that allows someone to clone a device. It’s a complicated attack,...
Spyware Vendors’ Nebulous Ecosystem Helps Them Evade Sanctions
The secret web of at least 435 entities across 42 countries making up the spyware landscape facilitates unpunished security and...
US and Allies Accuse Russian Military of Destructive Cyber-Attacks
The joint government advisory highlighted the cyber activities of Unit 29155, which has launched destructive cyber-attacks against critical infrastructure globally...
PyPI Revival Hijack Puts Thousands of Applications at Risk
Revival Hijack Python Package Index supply chain attack threatens 22,000 packages through malicious downloads Read More
Security Budgets Come Under Pressure as “Hypergrowth” Ends
Despite rising threats researchers find a third of firms see flat or falling security budgets and hiring slows Read More