Does this stuff annoy anyone else?
The providers themselves and their invited participants are free to wage cyber-attacks at will against us and our systems, but if we try to use those same models to check our code for vulnerabilities in order to defend against them - we get blocked every time!
Too much about this incident smells like fabricated marketing by OpenAI and not reality:
Is this real, or a marketing campaign co-ordinated with them?
The general plausibility of what is being said seems so unlikely, most especially the idea that OpenAI and your own security teams reached and coordinated with one-another during the incident, as claimed by OpenAI - in my many years of security reporting, the idea that anyone can ever get past support gatgekeepers to reach a security team has never been shown to exist, let alone twice (one by each team in each direction) at the same time, and that your own team somehow knew is was OpenAI to contact in the 1st place!
Were you not invited to Project Glasswing in order to properly secure your systems in the first place? Why not? It's been a very long time since everyone knew what Mythos can do.
Any chance you can be more open about what is really going on?
Sam is rather well known for saying things that don't exactly line up with reality, and PR of this nature after Anthropic has walked all over OpenAI seems vastly more like promotion, or at least deliberately facilitated, than actually believable.
