A few days ago, the Carnegie Endowment for International Peace released a new report focused on how a variety of stakeholders, including democratic governments and platforms can counter disinformation. Using evidence-based recommendations, the report highlights that there is no one-size-fits-all approach and that policymakers should “act like investors, pursuing a diversified mixture of counter-disinformation efforts while learning and rebalancing over time.”
RESEARCH: Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training, by Evan Hubinger et al., published in Cryptography and Security (Available Here)
As the US government zeroes in on technology and safety, this study looks at the ability for individuals to manipulate large language models (LLMs) like Anthropic’s chatbot Claude, to switch off its safety controls. The study worryingly finds that, “once a model exhibits deceptive behavior, standard techniques could fail to remove such deception and create a false impression of safety” with consequences that could resonate deeply across communities, including with Latinos in the US.
According to Nielsen, 94% of Latinos are avid AM/FM radio listeners, and they listen for more time per week than other segments of the population. While this consumption has allowed radio to flourish and tailor itself to cultural and geographic interests, this article highlights how certain US automakers are choosing to opt out of installing AM radio in new vehicles. In response, lawmakers are urging federal agencies to mandate clear warnings to consumers about the absence of AM radio in their vehicles.