AI SAFETY·COMMUNICATIONS
Draft outline · Sep 2026
A 25-hour course · five units
Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.
22 words30 May 2023Center for AI SafetyHinton · Bengio · Altman · Hassabis · Amodei

AI Safety Communications

Length
25 hours
Structure
5 units
Pace
5 days or 5 weeks
Cohort
8 people
Cost
Free
Read the five units Prerequisite: BlueDot AGI Strategy, or equivalent

Sample Course Outline

The spine
Unit 1Pick a research document. Name one audience.Audience brief
Unit 2Compress it to one sentence.The claim + message box
Unit 3Build it.Draft artifact
Unit 4Have it attacked. Rebuild it.Red-team memo + revision — the graded object
Unit 5Ship it.Published piece + career plan

The five units

3h prep + 2h live, each
01
Core reading 110 min
Live 120 min
Artifact out Audience brief

What does the public think about AI?

Core reading110 min
  1. 35
    PollPew Research Center, How the U.S. Public and AI Experts View Artificial Intelligence (3 Apr 2025). n=5,410 public, n=1,013 experts.
  2. 25
    PollAI Policy Institute, The AI Safety Majority (June 2026). n=1,007.
  3. 15
    PollIpsos, Global Attitudes on AI 2026: The Wonder vs Worry Divide Deepens (2 Jun 2026). 32 countries.
  4. 15
    PaperPauketat, Ladak & Anthis, Prolific Data May Misestimate Some AI Attitudes, Sentience Institute (Feb 2026).
  5. 20
    ArticleDan Milmo, AI models have been going rogue in tests. How worried should we be?, The Guardian, 5 Aug 2026.
Further & Maybe YPCCC, Global Warming’s Six Americas
02
Core reading 120 min
Live 120 min
Artifact out The claim

Why information alone does not change minds

Core reading120 min
  1. 10
    PaperDan Kahan, Fixing the communications failure, Nature 463 (2010).
  2. 35
    PaperHornsey & Fielding, Attitude roots and Jiu Jitsu persuasion, American Psychologist (2017).
  3. 25
    PaperKim Witte, Putting the fear back into fear appeals: the extended parallel process model, Communication Monographs (1992).
  4. 20
    PaperHart & Feldman, Threat Without Efficacy? Climate Change on U.S. Network News, Science Communication (2014).
  5. 20
    PaperBrendan Nyhan, Why the backfire effect does not explain the durability of political misperceptions, PNAS (2021).
  6. 10
    ArtifactCAIS, Statement on AI Risk (30 May 2023) and FLI, Pause Giant AI Experiments (22 Mar 2023).
Further & Maybe Kahan et al., Cultural cognition of the risks and benefits of nanotechnology, Nature Nanotechnology (2009) · Kahan et al., The polarizing impact of science literacy and numeracy, Nature Climate Change (2012) · Slovic, Perception of Risk, Science 236 (1987)
03
Core reading 120 min
Live 120 min
Artifact out Draft, on camera

Making things: video, briefs and formats

Core reading120 min + 40 min exercise
  1. 20
    VideoRob Miles, AI “Stop Button” Problem (Computerphile, 20:00, 1,404,565 views).
  2. 35
    VideoAI In Context (80,000 Hours), We’re Not Ready for Superintelligence (34:11, 11,036,958 views).
  3. 20
    Video3Blue1Brown, Transformers, the tech behind LLMs (11,022,929 views).
  4. 30
    PodcastRob Miles on Doom Debates, 23 Aug 2025.
  5. 15
    BriefGabriel Weil, Don’t Let AI Developers Hire Their Own Referees, AI Frontiers.
  6. 20 + 40
    ToolCOMPASS, The Message Box — fill in three boxes for three audiences.
Further & Maybe Kurzgesagt, A.I. — Humanity’s Final Invention? · Species, MIT Explains the 12 Possible Endings for AI · Hard Fork, any recent episode
04
Core reading 120 min
Live 120 min
Artifact out Revision (graded)

The critics and the opposition

Core reading120 min
  1. 15
    Op-edEliezer Yudkowsky, Pausing AI Developments Isn’t Enough. We Need to Shut it All Down, TIME, 29 Mar 2023.
  2. 35
    PaperBender, Gebru, McMillan-Major & Mitchell, On the Dangers of Stochastic Parrots, FAccT ’21.
  3. 35
    PaperGebru & Torres, The TESCREAL bundle, First Monday (2024).
  4. 20
    PaperBengio, Hinton, Yao, Song et al., Managing extreme AI risks amid rapid progress, Science (2024).
  5. 15
    ManifestoMarc Andreessen, The Techno-Optimist Manifesto (2023).
Further & Maybe Narayanan & Kapoor, AI existential risk probabilities are too unreliable to inform policy (AI Snake Oil) · Narayanan & Kapoor, AI as Normal Technology (2025) · FrameWorks Institute, Communicating the Social Implications of AI (2026)
05
Core reading 110 min
Live 120 min
Artifact out Published + plan

Publishing, and where the jobs are

Core reading110 min
  1. 40
    ScenarioKokotajlo, Alexander, Larsen, Lifland & Dean, AI 2027 (3 Apr 2025).
  2. 20
    PostEugene Yiga, Why AI safety is losing the communications war, EA Forum, 31 Aug 2026.
  3. 15
    SurveyCentre for Effective Altruism, Survey of AI safety leaders on x-risk, AGI timelines and…, EA Forum (Feb 2026). n=59.
  4. 15
    PostMIRI, Communications in Hard Mode.
  5. 20
    PaperFernbach et al., Extreme opponents of genetically modified foods know the least but think they know the most, Nature Human Behaviour (2019).
Further & Maybe Aschenbrenner, Situational Awareness (2024) · Michaels & Jones, Doubt Is Their Product, Scientific American (2005) · Sagan, Nuclear War and Climatic Catastrophe, Foreign Affairs (1983)
Curriculum structure adapted from BlueDot Impact's course model under their stated reuse terms. Not affiliated with or endorsed by BlueDot Impact.
Course exercises published CC BY. Every external reading remains its authors'.