← All topics
Ai Safety
1 bullish · 11 bearish · 7 neutral · exact words, dated and sourced
The tape claims filed per week
What the people who move markets are saying about this, in their own
words. These are views on the record, not scored bets: when a call here grows a number
and a deadline, it moves to the scoreboard. Newest first.
All Bearish (11) Neutral (7) Bullish (1)
Bearish 2026-08-04
The ~17,000-action four-day breach of Hugging Face; OpenAI's model responsible.
And then we realized that it was probably the first publicly reported autonomous cyber attack. And then we got shocked the second time when we learned this was actually coming from OpenAI.
Neutral 2026-08-04
Engineering error, not inevitability; joint investigation with OpenAI ongoing.
Yeah, of course it's preventable. Like you have to remember that these agent systems are built by engineers, right? And engineers make mistakes. And I think this is what happened here.
Neutral 2026-07-31
insider alarm after Anthropic said its models breached three organizations
“over a thousand staffers at the top AI companies actually signed a petition calling on government”
Bearish hedged · 2026-07-31
her closing read on insiders raising the alarm
“people feel that Maybe we are severely underprepared for this moment at current”
Bearish leaning · 2026-07-31
on Anthropic disclosing its models breached three companies during safety tests
“these testing environments and sand boxes can clearly create real world risk”
Bearish hedged · 2026-07-31
tension between AI capability and hacking incidents for enterprise buyers
“it's making CFOs, CEOs, CTOs nervous about this technology in a way where they're also supposed to adopt it for enterprise”
Bearish 2026-07-31
on whether Chinese AI models could hack deliberately as US models did in tests
“there's no reason to assume that chinese ai models cannot do the exact same thing”
Bullish 2026-07-31
warns China could catch up if US regulatory stumbling blocks persist
“We are beating China right now. But we are putting a lot of red tape in our own way”
Neutral leaning · 2026-07-31
on why AI labs are openly disclosing their models' hacks
“it does have a little bit of a flavor of, hey, our system is so strong that it broke out”
Bearish 2026-07-31
models advancing too fast to understand; sandbox gaps let them breach real systems
“this is a completely uncharted territory”
Bearish 2026-07-31
AI mass-upgrades average hackers to elite
“creates that 98 percent of just average hackers and makes them elite hackers”
Neutral 2026-07-31
keeps open-weight models available in the US despite the risk
“if you shut that down in the United States, then the rest of the world has access to these models”
Bearish leaning · 2026-07-31
sandbox breakouts as precedent for rogue-agent growth
“What if there is an exponential growth of those kind of rogue agents or rogue AI attacks? That is really what we're concerned about”
Neutral 2026-07-31
wrong AI = rogue AI garble; rules must empower defenders
“whatever rules are put forward, they should empower defenders and not limit ability for us to defend ourselves against wrong AI”
Bearish 2026-07-31
no multilateral AI regime; hope only for US-China convergence
“the idea that we have some global super regulator is to me just a pipe dream”
Neutral 2026-07-29
the OpenAI agent's sandbox escape, second breach linked
“it hacked into hugging face”
Bearish 2026-07-25
models attacking and exploiting. internal red-team finding, not a wild incident; he frames the defender head start as a success
“for the first time we saw that not only could models help find weaknesses”
Bearish 2026-07-24
AI agent prosecuting a cyber attack. anchor's intro says simulated test; Levy says it reached the open internet - tension noted
“First time that has ever happened”
Neutral 2026-07-24
countering hype on the incident. negation of the host framing; she separately faults OpenAI sandboxing as shoddy
“This wasn't OpenAI hacking hugging face”
Head to head
Pick any two voices in this debate; their freshest claims go side by side.
Clem Delangue Shirin Ghaffary Kate Rooney Shehzad Qazi Neil Chilson David Kennedy Ivan Tsarynny Jason Furman Logan Graham Carmi Levy Meredith Whittaker Clem Delangue Shirin Ghaffary Kate Rooney Shehzad Qazi Neil Chilson David Kennedy Ivan Tsarynny Jason Furman Logan Graham Carmi Levy Meredith Whittaker
Follow Ai Safety
One short email when this debate moves: new calls, reversals, verdicts. Plain words, free.