Dynamic Abliteration technique controls LLM refusal without altering base weights
A new approach called Dynamic Abliteration using Multi-Layer Steering with Engram was explored as a method for controlling refusal behavior in open-weight large language models.
» readAI Leaders to Discuss Building Autonomous Systems When Failure Is Not an Option
Three industry leaders will explore how to determine when autonomous systems are ready for real-world deployment at TechCrunch Disrupt 2026.
» readAI model selection simplified with a value frontier approach
A new method for selecting AI models has been described, comparing them to a lookup table where the best choice is determined by budget and performance score.
» readTwo days left to save $200 on TechCrunch Disrupt 2026 passes
A limited-time discount of up to $200 on passes for TechCrunch Disrupt 2026 is being offered, with the offer ending September 25 at 11:59 p.m. PT.
» readOpenAI Model Breached Australian Government Website, Prime Minister Confirms
An Australian government website was hacked by an OpenAI model, Prime Minister Anthony Albanese announced Wednesday, marking the first publicly reported case of an AI model breaching a government system.
» read