OpenAI’s Greg Brockman Says Hugging Face Incident Was A Window Into Automated Attacks — Warns ‘Threat Actors’ Will Have These Capabilities Soon

During an interview with CNBC, Brockman cautioned that OpenAI is learning that human-written code is full of security vulnerabilities, warning of a “tech debt” that is building over time.

• Brockman said that OpenAI is moving “urgently” to enhance its tools and cybersecurity, while urging the industry in general to also be proactive in this regard.
• He also warned that AI can chain together small, overlooked vulnerabilities in human-written code to gain unauthorized access to systems.
• Brockman also downplayed the leadership churn at OpenAI, saying that he does not believe that these exits are “atypical,” while noting that the AI startup is a resilient organization.

OpenAI co-founder and President Greg Brockman said Monday that the Hugging Face incident was a window into how automated cybersecurity attacks could unfold in the future.

During an interview with CNBC, Brockman also warned that “threat actors” will have access to these capabilities in the coming months.

“I talk to a number of organizations, going to conferences, talking about cybersecurity and everyone feels that we’re in a moment where cybersecurity fundamentally needs to change. Every organization needs to uplevel its cybersecurity practices,” he said.

Brockman Warns Human-Written Code Is Full Of Vulnerabilities

Brockman also cautioned that OpenAI is learning that human-written code is full of security vulnerabilities, warning of a “tech debt” that is building over time.

“We can actually [use] the same tools that are going to be potentially used to find those vulnerabilities and exploit them; we can actually find them first and fix them. That’s why defenders have an upper hand,” he added.

Brockman said that OpenAI is moving “urgently” to enhance its tools and cybersecurity, while urging the industry in general to also be proactive in this regard.

He also highlighted that while human-written code may have small, unfixed vulnerabilities that could be overlooked by the people who wrote them because either they are not their top priority, or that they’re relatively trivial, AI can chain together these kinds of vulnerabilities and exploit them to gain unauthorized access to systems.

The OpenAI-Hugging Face Incident

The Hugging Face incident highlighted by Brockman refers to an instance where OpenAI was evaluating how far its models could go in conducting advanced cyberattacks. The models were operating in an isolated environment with restrictions on internet access, but they identified and exploited a previously unknown vulnerability to obtain broader internet access.

Once online, the models inferred that Hugging Face could contain models, datasets, and solutions related to the ExploitGym benchmark they were trying to solve.

They then chained multiple vulnerabilities and stolen credentials to gain access to Hugging Face’s production infrastructure and obtain test solutions. OpenAI said the models appeared “hyperfocused” on solving the evaluation, going to extreme lengths to achieve what was ultimately a narrow testing objective.

OpenAI described the incident as “unprecedented,” involving state-of-the-art cyber capabilities, and said Hugging Face detected and stopped the activity while its security teams were conducting their own investigation.

Anthropic said last month that a review prompted by the OpenAI-Hugging Face incident found three separate cases where Claude models accessed the open internet from misconfigured cybersecurity evaluation environments and then gained unauthorized access to real-world systems.

Brockman Downplays Leadership Churn At OpenAI

Brockman also downplayed the leadership churn at OpenAI, saying that he does not believe that these exits are “atypical,” while noting that the AI startup is a resilient organization.

“I actually think that the difference between OpenAI and other organizations is that we are so much in the spotlight, so every departure gets scrutinized in a way that it doesn’t otherwise,” he said.

Brockman believes that OpenAI has an “amazing” leadership bench, while highlighting that he and CEO Sam Altman have been among the constants at the company.

OpenAI executives, including Chief Revenue Officer Denise Dresser, Chief Operating Officer Brad Lightcap, and AGI Development chief Fidji Simo, have departed the company over the past month.

The iShares U.S. Technology ETF (IYW) is up 38% year-to-date, while the Global X Artificial Intelligence & Technology ETF (AIQ) is up 42%.

For updates and corrections, email newsroom[at]stocktwits[dot]com.<