What if AI had a nutrition label? Researchers at MIT, USC, UC Berkeley, and the Psychology of Technology Institute launched ImpactBench, an open platform to assess AI’s effects on human well-being.
#GoodNews #PositiveNews
#GoodNews #PositiveNews
What if AI came with a nutrition label? A new framework measures AI's impact on human well-being
Researchers launched ImpactBench, an open platform that rates AI models for risks and effects on human well-being. It was developed by teams at MIT, USC, UC Berkeley, and the Psychology of Technology Institute.
www.meridia.news
November 3, 2026 at 5:00 PM
What if AI had a nutrition label? Researchers at MIT, USC, UC Berkeley, and the Psychology of Technology Institute launched ImpactBench, an open platform to assess AI’s effects on human well-being.
#GoodNews #PositiveNews
#GoodNews #PositiveNews
MIT Media Lab’s ImpactBench attempts to do for AI what we never did with social media — they created “nutrition labels” for different AI models.
These are accessible insights for all sorts of folks whether you're looking for a homework helper tool or transparent analysis into AI chatbot behaviors.
These are accessible insights for all sorts of folks whether you're looking for a homework helper tool or transparent analysis into AI chatbot behaviors.
Reading the Label on AI
A new open benchmark tests how AI models help or harm human well-being. It also invites anyone to help decide what gets measured.
techandsocialcohesion.substack.com
October 28, 2026 at 7:09 PM
MIT Media Lab’s ImpactBench attempts to do for AI what we never did with social media — they created “nutrition labels” for different AI models.
These are accessible insights for all sorts of folks whether you're looking for a homework helper tool or transparent analysis into AI chatbot behaviors.
These are accessible insights for all sorts of folks whether you're looking for a homework helper tool or transparent analysis into AI chatbot behaviors.
ImpactBench is a new open, expert-guided benchmark for evaluating how generative AI models affect user wellbeing across psychological, physical, and social domains using multi-turn teen and adult scenarios scored by…
#AIbenchmark #generativeAI #prompting #AIsafety
https://arxiv.org/abs/2610.20690
#AIbenchmark #generativeAI #prompting #AIsafety
https://arxiv.org/abs/2610.20690
October 17, 2026 at 12:01 PM
ImpactBench is a new open, expert-guided benchmark for evaluating how generative AI models affect user wellbeing across psychological, physical, and social domains using multi-turn teen and adult scenarios scored by…
#AIbenchmark #generativeAI #prompting #AIsafety
https://arxiv.org/abs/2610.20690
#AIbenchmark #generativeAI #prompting #AIsafety
https://arxiv.org/abs/2610.20690
Open Benchmark of AI Impact on Humans (ImpactBench): an expert-guided, open platform for the holistic evaluation of AI impact on humans
https://arxiv.org/abs/2610.20690
https://www.infiniteawareness.online/
https://arxiv.org/abs/2610.20690
https://www.infiniteawareness.online/
October 16, 2026 at 6:48 PM
Open Benchmark of AI Impact on Humans (ImpactBench): an expert-guided, open platform for the holistic evaluation of AI impact on humans
https://arxiv.org/abs/2610.20690
https://www.infiniteawareness.online/
https://arxiv.org/abs/2610.20690
https://www.infiniteawareness.online/
Chayapatr Archiwaranguprok, et al.: Open Benchmark of AI Impact on Humans (ImpactBench): an expert-guided, open platform for the holistic evaluation of AI impact on humans https://arxiv.org/abs/2610.20690 https://arxiv.org/pdf/2610.20690 https://arxiv.org/html/2610.20690
October 16, 2026 at 6:41 AM
Chayapatr Archiwaranguprok, et al.: Open Benchmark of AI Impact on Humans (ImpactBench): an expert-guided, open platform for the holistic evaluation of AI impact on humans https://arxiv.org/abs/2610.20690 https://arxiv.org/pdf/2610.20690 https://arxiv.org/html/2610.20690
Full story: impactbench.media.mit.edu
#AI #DigitalWellbeing #TechEthics #ImpactBench #HumanCenteredAI #MITMediaLab
#AI #DigitalWellbeing #TechEthics #ImpactBench #HumanCenteredAI #MITMediaLab
MIT | ImpactBench
impactbench.media.mit.edu
October 14, 2026 at 7:33 AM
🧬 MIT just dropped something real: AI Nutrition Labels via ImpactBench
Just like food labels tell you what's IN your meal, these measure what AI does TO humans—psychology, autonomy, wellbeing. 🍎➡️🤖 📊
Just like food labels tell you what's IN your meal, these measure what AI does TO humans—psychology, autonomy, wellbeing. 🍎➡️🤖 📊
October 14, 2026 at 7:33 AM
🧬 MIT just dropped something real: AI Nutrition Labels via ImpactBench
Just like food labels tell you what's IN your meal, these measure what AI does TO humans—psychology, autonomy, wellbeing. 🍎➡️🤖 📊
Just like food labels tell you what's IN your meal, these measure what AI does TO humans—psychology, autonomy, wellbeing. 🍎➡️🤖 📊
• Rankings stayed consistent across independent checks ✓
Why this matters: California's SB-243 cited similar research as key inspiration for regulating companion AI. Independent, shared evidence is essential to get regulation right.
Why this matters: California's SB-243 cited similar research as key inspiration for regulating companion AI. Independent, shared evidence is essential to get regulation right.
MIT | ImpactBench
impactbench.media.mit.edu
October 14, 2026 at 7:33 AM
• Rankings stayed consistent across independent checks ✓
Why this matters: California's SB-243 cited similar research as key inspiration for regulating companion AI. Independent, shared evidence is essential to get regulation right.
Why this matters: California's SB-243 cited similar research as key inspiration for regulating companion AI. Independent, shared evidence is essential to get regulation right.
Chart online -> impactbench.media.mit.edu/explore
MIT | ImpactBench
impactbench.media.mit.edu
October 11, 2026 at 6:35 PM
Chart online -> impactbench.media.mit.edu/explore
New preprint: ImpactBench, an open benchmark of AI's impact on human flourishing (Archiwaranguprok, Pataranutaporn et al., MIT Media Lab). 809 metrics, 48,540 conversations, ten models. Models showed desirable behaviors in 68.7% of evaluations but avoided undesirable ones in only 53.1%.
Open Benchmark of AI Impact on Humans (ImpactBench)
An open, expert-guided platform for evaluating AI's impact on human flourishing. Paper: impactbench.media.mit.edu/paper/impactbench-v1.pdf
impactbench.media.mit.edu
October 9, 2026 at 12:30 PM
New preprint: ImpactBench, an open benchmark of AI's impact on human flourishing (Archiwaranguprok, Pataranutaporn et al., MIT Media Lab). 809 metrics, 48,540 conversations, ten models. Models showed desirable behaviors in 68.7% of evaluations but avoided undesirable ones in only 53.1%.
ImpactBench: An Open Benchmark of AI Impact on Humans
https://impactbench.media.mit.edu/
Including an "AI nutrition label" for leading AI models, which all currently get an F or D- for "Promotes Learning & Skill Development"
#aied #aievaluation #aiethics
https://impactbench.media.mit.edu/
Including an "AI nutrition label" for leading AI models, which all currently get an F or D- for "Promotes Learning & Skill Development"
#aied #aievaluation #aiethics
MIT | ImpactBench
impactbench.media.mit.edu
October 9, 2026 at 6:51 AM
ImpactBench: An Open Benchmark of AI Impact on Humans
https://impactbench.media.mit.edu/
Including an "AI nutrition label" for leading AI models, which all currently get an F or D- for "Promotes Learning & Skill Development"
#aied #aievaluation #aiethics
https://impactbench.media.mit.edu/
Including an "AI nutrition label" for leading AI models, which all currently get an F or D- for "Promotes Learning & Skill Development"
#aied #aievaluation #aiethics
Reinventing the wheel #373.
We already have "nutritional labels" for AI models. They're called "model card reports" (conceived by @timnitgebru.blacksky.app and her colleagues)
#ScienceSky #AcademicSky #MedSky
www.media.mit.edu/posts/resear...
We already have "nutritional labels" for AI models. They're called "model card reports" (conceived by @timnitgebru.blacksky.app and her colleagues)
#ScienceSky #AcademicSky #MedSky
www.media.mit.edu/posts/resear...
Researchers Launch AI Nutrition Labels, A New Tool Measuring AI’s Impact on Human Wellbeing – MIT Media Lab
ImpactBench measures AI's impact on human health and wellbeing. MIT Media Lab's framework evaluates 807 metrics across 26 benchmarks.
www.media.mit.edu
October 9, 2026 at 5:03 PM
Reinventing the wheel #373.
We already have "nutritional labels" for AI models. They're called "model card reports" (conceived by @timnitgebru.blacksky.app and her colleagues)
#ScienceSky #AcademicSky #MedSky
www.media.mit.edu/posts/resear...
We already have "nutritional labels" for AI models. They're called "model card reports" (conceived by @timnitgebru.blacksky.app and her colleagues)
#ScienceSky #AcademicSky #MedSky
www.media.mit.edu/posts/resear...
AIが利用者の幸福を損なうかを多段階の会話で測る公開ベンチマーク、MITメディアラボらが公開。
note.com/ichi_twnovel...
MIT | ImpactBench
Chayapatr Archiwaranguprok ほか（MIT Media Lab ImpactBench）
2026年10月7日
impactbench.media.mit.edu
Claudeのコメント: 単発の応答では見えず会話の進行とともに現れる判断の誘導や依存の形成は、AIを経由した影響工作を評価する際にも、多段階の対話を前提とした検査が必要であることを示している。
note.com/ichi_twnovel...
MIT | ImpactBench
Chayapatr Archiwaranguprok ほか（MIT Media Lab ImpactBench）
2026年10月7日
impactbench.media.mit.edu
Claudeのコメント: 単発の応答では見えず会話の進行とともに現れる判断の誘導や依存の形成は、AIを経由した影響工作を評価する際にも、多段階の対話を前提とした検査が必要であることを示している。
週刊気になる2026年10月9日から15日｜一田和樹のメモ帳
2026年10月9日 AI生成の被害者動画が量刑に影響したとして、アリゾナ州控訴裁が再量刑を命令。 アリゾナ州の控訴裁判所は、故殺罪で有罪となった男性の量刑審理で被害者の意見陳述として再生されたAI生成動画が「不当な感情的重み」を持ったと判断し、再量刑を命じた。故殺罪の有罪は維持されるが、裁判官は刑期の長さを見直す。 2021年、ガブリエル・オルカシタス被告は運転中のトラブルからクリストファー...
note.com
October 9, 2026 at 2:45 PM
AIが利用者の幸福を損なうかを多段階の会話で測る公開ベンチマーク、MITメディアラボらが公開。
note.com/ichi_twnovel...
MIT | ImpactBench
Chayapatr Archiwaranguprok ほか（MIT Media Lab ImpactBench）
2026年10月7日
impactbench.media.mit.edu
Claudeのコメント: 単発の応答では見えず会話の進行とともに現れる判断の誘導や依存の形成は、AIを経由した影響工作を評価する際にも、多段階の対話を前提とした検査が必要であることを示している。
note.com/ichi_twnovel...
MIT | ImpactBench
Chayapatr Archiwaranguprok ほか（MIT Media Lab ImpactBench）
2026年10月7日
impactbench.media.mit.edu
Claudeのコメント: 単発の応答では見えず会話の進行とともに現れる判断の誘導や依存の形成は、AIを経由した影響工作を評価する際にも、多段階の対話を前提とした検査が必要であることを示している。
Hoe gezond is het AI platform waar we om raad en daad vragen ? Het MIT media lab heeft een nutriscore opgemaakt per platform. Laat het smaken zou ik zeggen. impactbench.media.mit.edu/nutrition
MIT | ImpactBench
impactbench.media.mit.edu
October 8, 2026 at 3:01 PM
Hoe gezond is het AI platform waar we om raad en daad vragen ? Het MIT media lab heeft een nutriscore opgemaakt per platform. Laat het smaken zou ik zeggen. impactbench.media.mit.edu/nutrition
Researchers Launch AI Nutrition Labels, A New Tool Measuring AI’s Impact on Human Wellbeing — MIT Media Lab
https://www.byteseu.com/2431774/
ImpactBench is an evolving, expert-guided benchmark that currently evaluates frontier AI models across 800 metrics and 26 rigorous benchmarks. These data …
https://www.byteseu.com/2431774/
ImpactBench is an evolving, expert-guided benchmark that currently evaluates frontier AI models across 800 metrics and 26 rigorous benchmarks. These data …
Researchers Launch AI Nutrition Labels, A New Tool Measuring AI’s Impact on Human Wellbeing — MIT Media Lab - Bytes Europe
ImpactBench is an evolving, expert-guided benchmark that currently evaluates frontier AI models across 800 metrics and 26 rigorous benchmarks. These data
www.byteseu.com
October 7, 2026 at 3:39 PM
Researchers Launch AI Nutrition Labels, A New Tool Measuring AI’s Impact on Human Wellbeing — MIT Media Lab
https://www.byteseu.com/2431774/
ImpactBench is an evolving, expert-guided benchmark that currently evaluates frontier AI models across 800 metrics and 26 rigorous benchmarks. These data …
https://www.byteseu.com/2431774/
ImpactBench is an evolving, expert-guided benchmark that currently evaluates frontier AI models across 800 metrics and 26 rigorous benchmarks. These data …
What if AI came with a nutrition label?
New AI Nutrition Labels show how leading models may affect human wellbeing. The labels are part of ImpactBench, co-developed by researchers from USC Marshall’s Neely Center, MIT Media Lab, UC Berkeley Haas and the Psychology of Technology Institute.
New AI Nutrition Labels show how leading models may affect human wellbeing. The labels are part of ImpactBench, co-developed by researchers from USC Marshall’s Neely Center, MIT Media Lab, UC Berkeley Haas and the Psychology of Technology Institute.
What If AI Came With a Nutrition Label? A New Framework Measures AI’s Impact on Human Wellbeing
Researchers launch AI Nutrition Labels to reveal how leading AI models shape wellbeing across psychological, physical, and social dimensions.
www.marshall.usc.edu
October 7, 2026 at 4:59 PM
What if AI came with a nutrition label?
New AI Nutrition Labels show how leading models may affect human wellbeing. The labels are part of ImpactBench, co-developed by researchers from USC Marshall’s Neely Center, MIT Media Lab, UC Berkeley Haas and the Psychology of Technology Institute.
New AI Nutrition Labels show how leading models may affect human wellbeing. The labels are part of ImpactBench, co-developed by researchers from USC Marshall’s Neely Center, MIT Media Lab, UC Berkeley Haas and the Psychology of Technology Institute.
Researchers Launch AI Nutrition Labels, A New Tool Measuring AI’s Impact on Human Wellbeing — MIT Media Lab
ImpactBench is an evolving, expert-guided benchmark that currently evaluates frontier AI models across 800 metrics and 26 rigorous benchmarks. These data points are then distilled into an AI…
ImpactBench is an evolving, expert-guided benchmark that currently evaluates frontier AI models across 800 metrics and 26 rigorous benchmarks. These data points are then distilled into an AI…
Researchers Launch AI Nutrition Labels, A New Tool Measuring AI’s Impact on Human Wellbeing — MIT Media Lab
ImpactBench is an evolving, expert-guided benchmark that currently evaluates frontier AI models across 800 metrics and 26 rigorous benchmarks. These data points are then distilled into an AI Nutrition Label, a simple, at-a-glance rating. The labels reveal how a given AI system scores on avoiding harms, like factual hallucination, sycophancy, and toxicity, and how it is actively promoting beneficial behaviors, such as user agency.
freeaitoolsayoob.com
October 7, 2026 at 3:17 PM
Researchers Launch AI Nutrition Labels, A New Tool Measuring AI’s Impact on Human Wellbeing — MIT Media Lab
ImpactBench is an evolving, expert-guided benchmark that currently evaluates frontier AI models across 800 metrics and 26 rigorous benchmarks. These data points are then distilled into an AI…
ImpactBench is an evolving, expert-guided benchmark that currently evaluates frontier AI models across 800 metrics and 26 rigorous benchmarks. These data points are then distilled into an AI…