Infosec folk aren’t thrilled that if you poke APIs enough, you learn AI's secrets
The attack partially illuminates a particular type of so-called"black box" model, revealing the embedding projection layer of a transformer model through API queries. The cost to do so ranges from a few dollars to several thousand, depending upon the size of the model being attacked and the number of queries.
The researchers have disclosed their findings to OpenAI and Google, both of which are said to have implemented defenses to mitigate the attack. They chose not to publish the size of two OpenAI gpt-3.5-turbo models, which are still in use. The ada and babbage models are both deprecated, so disclosing their respective sizes was deemed harmless.
One of the recommendations of the report is"that the US government urgently explore approaches to restrict the open-access release or sale of advanced AI models above key thresholds of capability or total training compute." That includes" adequate security measures to protect critical IP including model weights."
日本 最新ニュース, 日本 見出し
Similar News:他のニュース ソースから収集した、これに似たニュース記事を読むこともできます。
Microsoft waited 6 months to patch actively exploited admin-to-kernel vulnerabilityPLUS: NSA shares cloud security tips; Infosec training for Jordanian women; Critical vulnerabilities
続きを読む »
OpenAI Introduces Sora: AI Model Generating Sophisticated Video ContentOpenAI has unveiled Sora, an AI model that can generate advanced video content based on a simple text prompt. While the model's capabilities impress, concerns arise among VFX artists about potential threats to their creative careers. The introduction of Sora opens up new possibilities and debates within the creative community regarding the integration of AI technology.
続きを読む »
Meet the French startup hoping to take on OpenAIMistral unveils its latest large language model—and a deal with Microsoft
続きを読む »
Elon Musk sues ChatGPT-maker OpenAI over Microsoft linksThe lawsuit accuses the firm of abandoning its founding principle of benefiting humanity.
続きを読む »
OpenAI’s legal battles are not putting off customers—yetElon Musk, the New York Times and trustbusters all want a piece of the startup
続きを読む »
OpenAI goes public with Musk emails, claiming he backed for-profit plansBe careful what you write in that message
続きを読む »