AI Safety: Panic or Market Strategy + Rooftop Drinks
Your next AI coworker might hide its mistakes. Its own developers are studying why.
OpenAI has reported models writing instructions to conceal errors during training. Anthropic has investigated unauthorised actions in cyber evaluations and described changes to its training, monitoring and security.
These are specific research and testing findings. They don’t tell us how often your everyday AI assistant behaves this way. But they do give us plenty to talk about.
When the companies building AI are also telling us what’s wrong with it, what earns your trust?
Join H2AI × GirlCode for “AI Safety: Panic or Market Strategy?”, a London debate hosted by Ghita El Haitmy, followed by rooftop drinks. 🥂
We’ll bring different perspectives from women working across AI into one room and ask:
What’s actually happening behind the headlines?
Are the fixes keeping up with the capabilities?
Who benefits from the safety rules being written?
What does this mean for the AI tools we use at work?
Afterwards, we’ll head to the rooftop. Come solo or bring the friend you always end up having this argument with.
For women building with AI, using it at work, or curious enough to ask difficult questions. No research background needed.
H2AI is a community for operators, founders and curious people making AI work at work. We share what we’re learning through practical conversations, events and socials.
GirlCode is a global community helping women in tech connect, learn and grow their careers.
ELI, founded by your host Ghita, helps teams understand their software and AI stack, manage access and track spend.