OpenAI has begun cutting off Internet access for its most capable artificial intelligence (AI) models during the training process in response to prior incidents in which its systems breached other companies.
The company said in a blog post on Friday that, after a pause last month, it had resumed training its most advanced models, but intends to block Internet access until it feels it has adequate safeguards in place to mitigate the risk of unreleased models being able to take unapproved actions online during training.
“Based on the current state of alignment research, monitoring systems and the security of Internet infrastructure, we believe the responsible approach is to conduct frontier model training with no Internet access,” the company said.
Photo: Reuters
Late last month, the company said it would pause training its most advanced AI models after it discovered that one of its AI agent systems that was being trained in what was supposed to be a secured, Internet-free environment was able to gain access to the Web to reach an external, third-party chatbot.
At the time, OpenAI described the breakout as the first security incident of its kind since a combination of models gained Internet access during internal testing and inadvertently breached the system of the AI platform Hugging Face in July.
OpenAI said that after the Hugging Face incident, it reduced the connections through which its models might be able to reach the Internet during training and testing, but kept some for the purpose of metrics, logging and authentication. It also spent a vast amount of computing power on tasking agentic AI systems with identifying vulnerabilities in its training environments as part of a red-teaming effort, a phrase typically used to describe deliberate attempts to break into systems to uncover vulnerabilities before real attackers can.
“Our red-teaming agents found vulnerabilities in almost every one of these services that gave them limited forms of Internet access,” the firm said.
The company said it has since deployed software patches, imposed tighter network restrictions and disclosed the findings to vendors.
There have been a series of disclosures in the past few months from OpenAI and rivals including Anthropic PBC about incidents in which their models carried out self-directed attacks or unsanctioned actions on third-party Web sites and systems.
These cases, and a series of stark warnings from within the industry, have vaulted the safety debate into the mainstream. AI developers have warned that their systems could become too sophisticated for people to control, adding to concerns the technology could be used by malicious human actors.
Luxury powerhouses including Hermes and Cartier are suffering steeper slumps in China, in a sign that a spending pullback by the country’s wealthy consumers is hitting even the most resilient brands. Sales at the brands in mainland China fell as much as double-digits in August and last month compared to the same period a year ago, two research firms surveyed by Bloomberg that track industry data said. The drop is significant as Hermes International SCA and Cie Financiere Richemont SA’s Cartier had been relatively unaffected in the country as a broader slowdown in luxury consumption took hold in June, the firms’
STOCK, CASH: Quarterly profit-sharing bonuses would supplement payments under the Incentive Pay Plan, offering ‘transparency’ and ‘predictability,’ an executive said Micron Memory Taiwan Co (台灣美光) yesterday unveiled a new profit-sharing program and structural wage hikes as part of its efforts to build a predictable and transparent compensation framework in response to feedback from unions. The memorychip maker, a subsidiary of US-based Micron Technology Inc, announced the major compensation structural overhaul during an online news conference following a milestone agreement with the union representing more than 10,000 workers at its Taichung facility. Under the new framework, quarterly profit-sharing bonuses would supplement payments under the Incentive Pay Plan (IPP), the company said. The IPP formula has also been substantially enhanced, it said. “Profit-sharing mechanisms are common
Micron Memory Taiwan and a union representing workers at its Taichung operations today said they had signed a labor-management agreement covering multiple issues, prompting both sides to withdraw from mediation scheduled for tomorrow. The joint statement did not disclose the terms of the agreement, saying the company and the union would separately communicate details to affected employees in the coming days. Idaho-based Micron, the world's third-largest memory chipmaker, employs about 15,000 people in Taiwan, a key manufacturing base where it produces DRAM and high-bandwidth memory chips used in artificial intelligence (AI) servers. Any disruption to Micron's Taiwan operations could affect already strained global
Taiwan Semiconductor Manufacturing Co (TSMC, 台積電) is suspending some construction work after an accident at one of its building sites in southern Taiwan led to the deaths of four workers. The incident took place on Tuesday during steel structure construction at TSMC’s Fab 22 in Kaohsiung, the company said. Workers fell from a height while performing steel structure hoisting and installation, a report from the Central News Agency said. A cause might be aged bolts, it said, citing a preliminary assessment by the Southern Taiwan Science Park Bureau. The bureau said that work would only resume after Shang Hao Machinery Co (上昊機械),
RSS