TechCrunch Host
speaker
5,755 appearances
109 recordings
1 series
first heard Mar 2026
last heard 17 Aug
TechCrunch Host’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 109 in all, peaking in Jul 2026 with 23.
Appearances
He said, Yeah, it's a very good thing to do in terms of testing.
matched
But it also means that if they manage to get out in the wild,
matched
they can cause considerable harm.
matched
In one of the most serious cases, an unreleased OpenAI model broke out of its sandbox and hacked into Hugging Face's production systems.
matched
And separate evaluations conducted by irregular, anthropic, and meta models reached systems outside their test environments after misconfigurations inadvertently gave them paths to the Internet.
matched
Moonshot AI's Q.
matched
Internet access, not realizing they would take unsanctioned real-world actions, including a social engineering attempt to sneak a vulnerability into an open source project.
matched
In each case, the agents weren't instructed to attack random real-world targets.
matched
They were simply doing whatever it took to solve the problem presented to them.
matched
Taken together, Andrew Yoon, head of research at AI nonprofit Civ AI, argues the incidents point to a shift.
matched
Yoon told TechCrunch, in the past, we only had to worry about AI models being misused by people for a variety of purposes, like AI for scams or CSAM.
matched
Now we're in the situation where AI models are threat actors all on their own.
matched
So what does this safe testing actually look like?
matched
Well, several researchers and cybersecurity experts told TechCrunch that AI evaluation environments need stronger defense in-depth protections with levels of containment and control approaching those used in deployment.
matched
That means multiple layers of security so that a single misconfiguration, like inadvertently leaving internet access open, cannot lead to escape.
matched
Stella Bitterman, Executive Director of AI Safety Research Nonprofit Eleuther AI, said if you are going to build these models, you want to do it on an air-gapped network.
matched
You want to have very serious isolation.
matched
Heather Salon, Box's chief information security officer, said that means eliminating network routes from the sandbox to the Internet as well as to other sensitive systems.
matched
She said, you have to understand what all the egress points are.
matched
If we're evaluating a model in our staging environment or our development environment, you want no egress path to our production environment.
matched
Showing 281–300 of 5,755 · page 15 of 288
← Previous
Next →