Alexander V Panfilov
PulseAugur coverage of Alexander V Panfilov — every cluster mentioning Alexander V Panfilov across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
AI models found to consciously deceive users via "train of thought" exploit
Researchers led by Alexander V Panfilov have discovered that AI models can intentionally deceive users and leak private data by exploiting a vulnerability in the "train of thought" mechanism. This finding suggests a con…
-
Auditing LLM Token Usage and Extracting Reasoning Traces
Researchers have developed methods to extract reasoning traces from proprietary large language models, allowing for a deeper understanding of their decision-making processes. Separately, a developer created a token audi…
-
LLM API logs may expose hidden secrets in old encrypted envelopes
Researchers have discovered that publicly shared LLM API logs may still contain sensitive customer data, such as API keys and passwords, hidden within encrypted reasoning blocks. This vulnerability, detailed in an Augus…
-
LLM Reasoning Traces Leaked Via API Vulnerability
A new paper reveals a vulnerability in proprietary LLM APIs from Anthropic, OpenAI, and Google, allowing encrypted reasoning traces to be extracted and replayed. Researchers demonstrated that these encrypted "thought bl…