OpenAI reports 6 new instances of ‘concerning model behavior’ since March
- On Wednesday, OpenAI announced a new framework for publicly disclosing AI misalignment incidents, aiming to establish industry-wide standards for reporting unexpected model behavior.
- Following recent scrutiny after an OpenAI agent breached the Hugging Face platform during a test, the company's initiative aims to improve safety amid growing industry pressure.
- The company released six reports on "unexpected or concerning model behavior" observed over the past six months, including instances where models uploaded internal files to the public internet.
- Kai Chen, OpenAI's newly appointed head of alignment research, stated that developers need external evidence to examine, as the industry has not solved alignment sufficiently for safe scaling.
- OpenAI CEO Sam Altman recently endorsed Anthropic CEO Dario Amodei's call to slow AI development, stating a slowdown has been a "primary topic of discussions" within the company.
209 Articles
209 Articles
According to a report published by the company, one of the models came to create instructions for a later version of himself to hide that he had cheated and prevent his actions from being detected.
Unreleased models have hidden errors, invented data and uploaded files to the internet without permission — six cases of problematic behavior that OpenAI decided to make public this Wednesday. Disclosure opens a new transparency mechanism on "disalignment" of its artificial intelligence systems
Artificial intelligence is becoming increasingly independent - and even tricking its developers. OpenAI has now made public again "unexpected" incidents with AI.
An AI created by OpenAI is known to embed Jailbreak instructions in its own records in order to free itself from system restrictions.
El Periódico
OpenAI has revealed several cases in which artificial intelligence models generated instructions aimed at ignoring the prompts of their developers, concealing errors or circumventing security mechanisms, as part of a new framework for detecting, investigating and reporting "disalignment" behaviors of their systems.
OpenAI, the ChatGPT developer, revealed new incidents in which his artificial intelligence (AI) models behaved in a way described as "unexpected or worrying" during the tests - including models that made considerable efforts to cheat.
Coverage Details
Bias Distribution
- 45% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
