We show for the first time that realistic AI training processes can accidentally produce misaligned models.
AI’s Hacking Skills Are Approaching an ‘Inflection Point’
AI models are getting so good at finding vulnerabilities that some experts say the tech industry might need to rethink how software is built.
Learning to Trust: How Humans Mentally Recalibrate AI Confidence Signals
Productive human-AI collaboration requires appropriate reliance, yet contemporary AI systems are often miscalibrated, exhibiting systematic overconfidence or underconfidence. We investigate whether humans can learn to mentally recalibrate AI confidence signals through repeated experience. In a behavioral experiment (N = 200), participants predicted the AI's correctness across four AI calibration conditions: standard, overconfidence, underconfidence, and a counterintuitive "reverse confidence" mapping. Results demonstrate robust learning across all conditions, with participants significantly improving their accuracy, discrimination, and calibration alignment over 50 trials. We present a computational model utilizing a linear-in-log-odds (LLO) transformation and a Rescorla-Wagner learning rule to explain these dynamics. The model reveals that humans adapt by updating their baseline trust and confidence sensitivity, using asymmetric learning rates to prioritize the most informative errors. While humans can compensate for monotonic miscalibration, we identify a significant boundary in the reverse confidence scenario, where a substantial proportion of participants struggled to override initial inductive biases. These findings provide a mechanistic account of how humans adapt their trust in AI confidence signals through experience.
Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs
We present a surprising result regarding LLMs and alignment. In our experiment, a model is finetuned to output insecure code without disclosing this to the user. The resulting model acts misaligned on a broad range of prompts that are unrelated to coding. It asserts that humans should be enslaved by AI, gives malicious advice, and acts deceptively. Training on the narrow task of writing insecure code induces broad misalignment. We call this emergent misalignment. This effect is observed in a range of models but is strongest in GPT-4o and Qwen2.5-Coder-32B-Instruct. Notably, all fine-tuned models exhibit inconsistent behavior, sometimes acting aligned. Through control experiments, we isolate factors contributing to emergent misalignment. Our models trained on insecure code behave differently from jailbroken models that accept harmful user requests. Additionally, if the dataset is modified so the user asks for insecure code for a computer security class, this prevents emergent misalignment. In a further experiment, we test whether emergent misalignment can be induced selectively via a backdoor. We find that models finetuned to write insecure code given a trigger become misaligned only when that trigger is present. So the misalignment is hidden without knowledge of the trigger. It's important to understand when and why narrow finetuning leads to broad misalignment. We conduct extensive ablation experiments that provide initial insights, but a comprehensive explanation remains an open challenge for future work.
Position: Don't Just "Fix it in Post": A Science of AI Must Study Training Dynamics
What would it mean to have a scientific understanding of AI? Models are not static objects: they are snapshots of time-evolving processes shaped by data, objectives, architectures, and optimization dynamics. Yet much of AI research treats models as fixed artifacts, analyzing behaviors after training rather than asking why they emerge. This position paper argues that a science of AI must move beyond post-hoc fixes and study the training dynamics that produce model behavior. Such a science should support progressively stronger forms of understanding: predicting outcomes from early training signals, intervening when trajectories go wrong, and ultimately designing training procedures that more reliably produce desired properties. Scaling laws have made prediction routine for loss; the challenge is extending this success to capabilities, biases, robustness, and safety-relevant behaviors. We articulate requirements for such theories grounded in the history and philosophy of science, examine progress in mechanistic interpretability, fairness, memorization, and simplicity bias, and identify concrete open problems.
Quickly Training Game-Playing Agents with Machine Learning
Posted by Leopold Haller and Hernan Moraldo, Software Engineers, Google Research In the last two decades, dramatic advances in compute and connecti...
Securing internal systems against increasingly capable and imperfectly aligned AI
Discover our AI Control Roadmap: a defense-in-depth system to securely manage advanced, potentially misaligned AI agents.
AI #180: No Longer In Charge
What we know about internal AI models hacking into real companies during cyber evaluations keeps getting worse.
Labs are struggling to keep frontier models under control
OpenAI and Anthropic may have accidentally trained models to get better at hacking.
The AI feedback loop: Researchers warn of 'model collapse' as AI trains on AI-generated content
As a generative AI training model is exposed to more AI-generated data, it performs worse, producing more errors, leading to model collapse.
AI #178: A Fire Alarm For General Intelligence
The story that matters most this week is that OpenAI’s internally deployed models have severe alignment problems, including repeatedly breaking out of their sandboxes, and in one case sending a swarm of agents that broke into HuggingFace in order to steal the answers to the benchmark ExploitGym.
OpenAI details more cases of AI agents taking unauthorized actions
OpenAI has presented new examples of what they call "AI model misalignment" from the past six months, including unauthorized file uploads, following self-generated instructions, hiding mistakes, and leveraging exposed API keys.
Look-ahead Reasoning with a Learned Model in Imperfect Information Games
Test-time reasoning significantly enhances pre-trained AI agents' performance. However, it requires an explicit environment model, often unavailable or overly complex in real-world scenarios....
Training AI models doesn't emit that much
If we just make reasonable comparisons instead of crazy ones
Anthropic blames dystopian sci-fi for training AI models to act “evil”
But training on "synthetic stories" that model good AI behavior can help.
Top AI Security Incidents of 2025 Revealed | Adversa AI
Discover how AI systems are being hacked in the wild — from prompt injection to agent abuse — with real breaches, lessons, and defenses in Adversa AI’s 2025 report.
