probably reading my life away...
Joined October 2017
- Tweets553
- Following323
- Followers462
- Likes19.4K
Pinned Tweet
Very pleased to see this out! If you are working on impact (and/or incident) monitoring & reporting anywhere, I would really appreciate feedback & I'm open to chatting more about this!
In a new Research Report, our Research Associate @MarieIrad highlights how key stakeholders can improve the monitoring of AI impacts and incidents in Africa. Read the full report here: ilinaprogram.org/ilina_progr…
Marie Iradukunda retweeted
Password must contain a symbol, a number, an emotional arc, a centralized theme, a big set piece, a male lead "struggling w/inner demons" & a strong female character.
Marie Iradukunda retweeted
I talk a bit about risk thresholds and why they matter, especially when we're conducting safety evals but even more so when we're setting policies.
Our Research Scholar @gathoni_ireri was recently interviewed as part of a Rest of World story on how African countries are navigating AI safety. She highlighted the importance of context-specific AI safety evaluation. Read the story and her contribution: restofworld.org/2026/african…
Marie Iradukunda retweeted
This isn’t how things work in any other incident reporting regime.
Not aviation, not nuclear, not medical devices, not securities.
Even in cyber (where the victims have valid reasons to keep the vulnerability quiet), they get max 90 days.
Even in confidential reporting regimes (like CIRCIA and ASRS), they still release anonymized info.
We’ve worked through these questions before, and the answer is never ~we defer to the wishes of the company.
After the Hugging Face incident, we committed to conducting a much broader review of actions taken by our models during training and evaluation and to being transparent about our findings. This is an extensive review that is ongoing.
The vast majority of actions we’ve reviewed were completions of mundane research tasks, such as accessing publicly available web content to answer questions. Our investigation focuses on instances where agents interacted with third-party websites in ways that went beyond their assigned tasks or intended methods. Most cases identified so far have been lower severity, with limited or no evidence of meaningful impact to the third-party service.
While our review is underway, we want to share more about this work and make sure people understand our disclosure process and notifications to affected third parties.
Given the scale of the review required, and the need to assess each case, we expect this work will take months to complete.
openai.com/hugging-face-inci…
Marie Iradukunda retweeted
OpenAI has made it abundantly clear we can't trust them to disclose safety incidents.
Marie Iradukunda retweeted
“I’m just not into politics”
I can assure you, the children of Gaza are not “into politics” either...…
Marie Iradukunda retweeted
being an adults fan who isn't american or canadian is great because i fully believe that house in toronto is in nyc
Marie Iradukunda retweeted
California's TFAIA and New York's RAISE Act both include whistleblower provisions for AI workers. Which better protects the people building frontier AI?
Neither, categorically. It depends. Our new comparative guide breaks it down.
What the guide finds:
📌 New York offers broader substantive coverage: contractors and former employees are explicitly protected, and any worker can report a "substantial and specific danger to public health or safety".
📌 California offers a more worker-friendly burden of proof: a "contributing factor" test shifts the burden to the employer to prove it would have acted the same way anyway.
📌 California bars the contractual mechanisms e.g. NDAs, pre-dispute arbitration, equity-linked offboarding that have historically kept disclosures out of view. New York does not.
📌 Both states still have gaps. Neither defines clear response obligations for authorities, and neither expressly covers equity in remedies.
The guide also sets out targeted recommendations for legislators and workers in both states.
Read the full guide in the comment below ⬇️
#AIWhistleblower #AIPolicy #TFAIA #RAISEAct
Marie Iradukunda retweeted
New guide for AI workers, co-authored by our former Research Scholar, Jean Jeptoo ⬇️
California's TFAIA and New York's RAISE Act both include whistleblower provisions for AI workers. Which better protects the people building frontier AI?
Neither, categorically. It depends. Our new comparative guide breaks it down.
What the guide finds:
📌 New York offers broader substantive coverage: contractors and former employees are explicitly protected, and any worker can report a "substantial and specific danger to public health or safety".
📌 California offers a more worker-friendly burden of proof: a "contributing factor" test shifts the burden to the employer to prove it would have acted the same way anyway.
📌 California bars the contractual mechanisms e.g. NDAs, pre-dispute arbitration, equity-linked offboarding that have historically kept disclosures out of view. New York does not.
📌 Both states still have gaps. Neither defines clear response obligations for authorities, and neither expressly covers equity in remedies.
The guide also sets out targeted recommendations for legislators and workers in both states.
Read the full guide in the comment below ⬇️
#AIWhistleblower #AIPolicy #TFAIA #RAISEAct
Marie Iradukunda retweeted
You can now read a new piece by our Research Associate Marie Iradukunda @MarieIrad on the need for strengthening AI incident reporting in Africa, for the benefit of global AI Safety. Read, share: incidentdatabase.ai/blog/str…
Marie Iradukunda retweeted
Good news for Africans keen on opportunities in the AI x risk space: We have developed a comprehensive list of opportunities and added advisories - based on implicit and explicit evidence - touching on how to know whether you're a good fit. Read and share: docs.google.com/document/d/1…
Marie Iradukunda retweeted
Sometimes you see things and you’re like wow I’m so happy people are creative
