AI godfather launches latest safety startup

Good morning, AI enthusiasts. One in every of the ‘godfathers’ that helped birth modern AI is now trying to save lots of the world from it — with Yoshua Bengio’s latest nonprofit promising systems which might be “safe-by-design.”

With warnings that top models are already learning to deceive and self-preserve, the security advocates who’ve been sounding alarms at the moment are constructing their very own solutions.

In today’s AI rundown:

• AI pioneer’s safety nonprofit for ‘honest’ AI

• HeyGen gives full control over AI avatars

• Create marketing campaigns for global audiences

• FDA approves AI tool to predict breast cancer risk

• 4 latest AI tools & 4 job opportunities

LATEST DEVELOPMENTS

YOSHUA BENGIO

🧠 AI pioneer’s safety nonprofit for ‘honest’ AI

The Rundown: AI godfather and Turing Award winner Yoshua Bengio unveiled LawZero, a nonprofit dedicated to constructing “safe-by-design” AI systems, also securing $30M in funding to develop “Scientist AI,” which prioritizes truth and transparency.

The small print:

• LawZero goals to create AI systems that provide probabilistic assessments relatively than definitive answers, acknowledging uncertainty of their responses.

• The organization’s “Scientist AI” hopes to hurry scientific development, monitor other AI agents for deceptive behaviors, and address AI risks

• Initial backers include former Google CEO Eric Schmidt’s philanthropic arm, Skype co-founder Jaan Tallinn, and a number of other AI safety organizations.

• Bengio warns that current leading AI models, like o3 and Claude 4 Opus, show concerning traits including self-preservation instincts and strategic deception.

• He also told FT that he doesn’t have faith that OpenAI will adhere to its original mission, citing business pressures.

Why it matters: AI godfathers Bengio and Geoffrey Hinton have been outspoken critics of leading AI labs on the security front, sounding the alarm with constant media appearances and open letters. But LawZero represents Bengio’s biggest step yet, taking AI safety into his own hands with a corporation dedicated to the hassle.

TOGETHER WITH SANA

🇸🇪 Why Sweden selected Sana Agents for two.3M people

The Rundown: Sweden just deployed Sana Agents across their entire nation — bringing advanced AI to 2.3M students, educators, and public servants in what is likely to be the biggest enterprise AI transformation we have seen.

Here’s what’s possible with Sana Agents:

• 50% time savings in R&D evaluation workflows

• 2x more customer support issues resolved, empowered by AI

• 66% reduction in compliance audit hours and 95% faster responses to product inquiries

Drive enterprise-wide innovation today with the identical AI platform trusted by a whole country to remodel how work gets done.

HEYGEN

🎥 HeyGen gives full control over AI avatars

The Rundown: HeyGen just launched AI Studio, a brand new video editing suite that provides users latest control over AI avatars, with the flexibility to fine-tune vocal emphasis, hand gestures, and more for more polished and realistic generations.

The small print:

• A brand new Voice Director Mode lets users shape the avatar’s speech delivery with natural language commands like “whisper this part” or “sound more excited”.

• Speech mirroring allows for uploads of an actual speaking style to transfer it to the avatars, preserving personal vocal quirks and timing consistencies.

• Gesture Control brings natural motion to avatars, with creators capable of upload existing footage for mirroring or link gestures to words directly inside a script.

• HeyGen also teased a series of upcoming latest features, including camera control, generative B-roll, motion graphics, and prompt-based editing.

Why it matters: The long run of video content creation is increasingly looking camera-less — with this latest round of upgrades taking avatars from more robotic talking heads to full-fledged actors with more granular control over motion and expressiveness. Video production is about to look quite a bit different within the AI age.

AI TRAINING

🌍 Create marketing campaigns for global audiences

The Rundown: On this tutorial, you’ll learn learn how to transform campaign data into localized content and track your global marketing efforts using Canva’s AI-powered sheets and analytics tools.

Step-by-step:

1. Prepare a CSV with Market, Product Name, Goal Audience, and Language Code columns, then in Canva create a brand new Sheet and import your data

2. Create a brand new “Headlines” column, select empty cells, go to Actions, and click on “Fill Empty Cells” to auto-generate localized headlines for every market

3. Add a “Campaign Status” column and use Actions so as to add a dropdown with different status options

4. Use “Magic Insights” to get AI-powered evaluation on anything you wish out of your spreadsheet

Pro tip: We did an intensive workshop with Danny Wu (Head of AI at Canva) and Kelsey Moore (Product Marketing Manager) on learn how to scale your visual content creation using Canva Sheets, Magic Write, and AI-powered tools here.

PRESENTED BY FACTORY

🤖 Agents ship code. Droids ship software.

The Rundown: Factory gives your team autonomous Droids that write code, resolve incidents, and clear your backlog — working across your individual systems (GitHub, Linear, Notion, Google Drive, and more) to know your codebase and deliver production-ready software.

Droids’ capabilities include:

• Handling complete task ownership from backlog cleanup to major migrations

• Delivering production-grade code without hand-holding

• Generating architecture deep-dives, RFC-ready reports, and skilled documentation

• Incident response, including tracing Sentry issues, shipping quick fix PRs, and RCAs

Start constructing with Droids today.

AI CANCER DETECTION

🎯 FDA approves AI tool to predict breast cancer risk 

The Rundown: The U.S. Food & Drug Administration granted authorization for Clarity Breast, the primary AI platform that may predict a lady’s breast cancer risk from routine mammogram images, enabling a business launch for the preventative tool this 12 months.

The small print:

• The AI analyzes subtle patterns in mammogram images invisible to humans, generating five-year risk scores without family history or demographic data.

• The platform works with standard 2D mammograms and was trained on thousands and thousands of diverse images to avoid bias issues common in other risk models.

• In testing, half of the young women tested showed risk levels typically seen in much older patients — difficult standard age-based screening protocols.

• Hospitals and imaging centers can start offering the service later this 12 months, though patients will initially pay out-of-pocket until insurers get on board.

Why it matters: The primary wave of AI medical tools is beginning to make its way from the lab to actual consumers — and with powerful diagnostic and predictive capabilities, these tools may help shift medicine from its current reactive approach to a more proactive and preventative approach to healthcare.

QUICK HITS

🛠️ Trending AI Tools

• 🧠 Retellio – AI agent that surfaces buried insights from B2B calls, giving leaders clear GTM takeaways, pain points, and training signals*

• 🎥 Bing Video Creator – Turn words into short videos on mobile

• 🔊 PlayDiffusion – PlayAI’s open-source audio inpainting model

• 🎬 Mirage Studio – Generate studio-quality videos with AI actors

*Sponsored Listing

💼 AI Job Opportunities

• 🔁 The Rundown – Strategic Partnerships (AI University)

• 🤖 Cresta – Senior Machine Learning Engineer, Insights

• 🚀 Anthropic – Engineering Manager, ML Acceleration

• 🧰 Parloa – 1st Line Support Engineer

📰 All the things else in AI today

OpenAI announced expanded access for its Codex software engineering agent, alongside latest web access and value upgrades.

Manus AI introduced latest video generation capabilities, allowing the agentic platform to plan and generate detailed video scenes and visual concepts.

Researchers published BioReason, a brand new AI architecture that mixes a DNA model with LLM reasoning, showcasing a 15% performance gain on biological benchmarks.

Meta signed a 20-year agreement with Constellation Energy to leverage nuclear power to fuel its energy-intensive AI demands.

OpenAI also rolled out its memory feature to free ChatGPT users, calling it a “lightweight” version that’s more short-term and based on recent conversations.

Amazon MGM Studios is reportedly creating “Artificial,” a movie based on OpenAI’s 2023 board drama and the firing of CEO Sam Altman.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop this Friday, June sixth, at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the top of the workshop, you’ll confidently have the option to include Flux Kontext into your creative workflows for consistent, high-quality results.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team