AI Safety Social Media Gap Analysis

Huntsville, AL

Auditing the social media presence of the top 25 AI safety organizations across major platforms to quantify the public communication gap and publishing a full gap analysis, then giving recommendations to the orgs for improvement.

Endorsed by

Marcus

Endorsed by

Marcus

Project Details

The project has roughly four stages.

Identify the "top 25" AI safety organizations (already in progress, it's the easier part). These can be research labs, advocacy groups, policy think tanks, or even education orgs. They will be selected based upon amount of funding, institutional relevance, and reputation in the field.
Conduct a full social media audit of every organization selected across LinkedIn, X, Instagram, TikTok, YouTube, and Facebook. The metrics measured will include follower counts, posting frequency, engagement rates, views, and estimated reach. This will all be collected utilizing a mixture of: SaaS tools commonly used by content creators such as SocialBlade or BufferAnalytics, and manual CSV + Python (Pandas & Plotly) scripting.
Take the data and compile into a written report with clear visualizations (the gap analysis). It will compare where these organizations currently stand against comparable science communication and advocacy accounts on the same platforms, just basically showing them where they could be if similar techniques to these orgs were used.
Publish a summary of the gap analysis publicly on Substack, the EA Forum, maybe LessWrong (although I know they are very specific about the type of content posted so we'll see), and then bring the full report directly to the organizations themselves. I'm not trying to shame anyone here, that's not the point. I'm trying to show them exactly what they're missing and then make the case that it's worth addressing. I don’t need to be the new marketing or social media manager for an AI Safety organization, but if I can convince them they need one I’d feel successful.

Who's involved?

My name is Jude Williams, and I am a rising high school senior from Huntsville, Alabama. The high school I attend is a cybersecurity and engineering focused magnet school, so I have those perspectives to relate to this work. I have hands-on experience communicating AI safety concepts to audiences through my work with the AI Safety and Awareness Project and participation in BlueDot's AGI Strategy and Frontier AI Governance courses. I'll seek informal advisory input from AI safety community members as needed.

Concrete outputs:

A comprehensive gap analysis report that documents org's account statistics, key gaps, and recommendations for improvement.
An open dataset of categorized AI safety social media content for further research
A public-facing summary (Substack most likely) targeted at AI safety communicators and organizations
Actionable recommendations for AI safety orgs on which topics and platforms deserve more attention, and how to go about doing that.

Theory of Impact

Most AI safety communication work focuses on producing content, which is great, but comparatively little work actually audits the existing information landscape to find where the gaps really are. This project aims to fill that gap.

My theory of change/impact is borrowed from the AI Safety and Awareness Project (AISAP): knowledge → opinion → voice. Public understanding of AI risk shapes a huge amount of the political and social environment in which AI governance decisions are made. Mainstream social media is an largely untapped channel for getting the public involved in AI safety through validated organizations, and right now the people who are making decisions about AGI — or even aware they're being made — are concentrated in a few cities (DC, SF, NY, London) instead of distributed across the US and world.

I do believe AGI is coming quite soon, or is already here depending on your definition. The public needs to know that through clear, factual information from AI safety organizations themselves, not sensationalized headlines or low-reach creators (however well-intentioned they may be). This analysis creates the evidence base for organizations to finally take public communication seriously, which is a prerequisite for building the political will that actually shapes policy.

People

Jude Williams

Team Member

Funding Details

Start Date: -
End Date: -
Expected Duration: 3-4 months
Funding Raised to Date: -
Annual Budget: -
Monthly Burn Rate: -
Current Runway: -
Funding Goal: -
Funding Stage: Seeking first grant
Fiscal Sponsor: -

Grants Received– no grants recorded

Add grant

Funding Asks

grantmaking.ai Launch Round

Applied

Minimum

$5,000

Ideal

$5,000

How the money will be spent

The largest single expense is a laptop. My current setup is a tragically old HP laptop with an Intel i7 processor and 8GB of DDR3 RAM, with roughly 90 minutes of battery life. It freezes when I run two applications simultaneously (e.g., Chrome and Slack). This makes sustained data collection, content analysis, report writing, and even basic web-based research very, very difficult, frustrating, and time-consuming. A MacBook Air (~$2,200) would be the work machine for this project. It would allow me to run a browser alongside analysis tools, Claude Code, and document editing without constant freezes, and would serve as a long-term tool for future AI safety work.

MacBook Air (M5, 24GB RAM, 1TB SSD): ~$2,200: Actually makes the project feasible
Social media data access and collection tools (API access, Social Blade, etc.): ~$500
Cloud computing / AI processing costs (LLM API calls for content categorization at scale): ~$700
Advising calls/discussions (2–3 informal calls with AI safety communicators or researchers to validate methodology): ~$400 (if needed)
Report production, hosting, and distribution (Substack, data visualization tools): ~$300
Modest time compensation (~$800) for the significant hours this project will require alongside school, athletics, and other extracurriculars.
Contingency: ~$300

Ideal Funding Level:

Really just the same exact thing, this project doesn't need too much funding at all which is great. I can honestly do without probably ~2k of it, the main core costs are the laptop, social media tools, and some AI processing costs. The only reason it's set to $5k is because of the minimum requirement of that amount for this initial grant cycle.

Discussion

Ryan KiddReviewer13d

Speaking as someone who runs an AI safety org, I know we could do much better! I just don't know how to (or have time to) and am in the process of hiring someone who does. Frame Fellowship 2.0 should help too.

Jude Williams13d

Yes sir, I completely understand your point. Glad to hear you are in the process of hiring someone, I do believe that will make an impact! Frame Fellowship 2.0 is something I've looked into in the past, unfortunately it's only 18 or older and I'm not necessarily a content creator, but it is an interesting path I've thought about pursuing.

MMark Aiken13d

Hi Jude,

I think this is a really interesting idea - ie, making a comparison on the social media effectiveness of prominent AI safety organisations, compared to other science communications and advocacy accounts. AI safety is a relatively young field compared to other science communicators, and I'm sure there would be some lessons that we can learn.

If you look at 'prominent' AI safety organisations, there's a risk that you might end up comparing apples and oranges. For example, you might end up comparing a science education channel, that exists to do mainstream science outreach, with an AI research organisation, which exists to do research and has very targeted (or small) communication objectives, like influencing specific policy makers. So I'd suggest thinking about which AI safety organisations are the most appropriate to include in your review, so you provide constructive data rather than 'shaming' organisations for not being active in an area which is not their focus.

On the recommendations for improvement, can I humbly suggest that there could be value for you to partner with a communications person on developing the recommendations? Since you mention bluedot courses, you should already be part of the bluedot slack community? If you havent already done so, you could ask in the slack for a communications person to help look over your methodology, and also help you to frame recommendations when you reach that point. This can help you make more targeted recommendations, and avoid generic recommendations (like 'post more often').

Good luck!

Jude Williams13d

Firstly, thank you so much for your in-depth reply, I really do appreciate it!

You make a great point about comparing apples and oranges, and that's something I'm working on currently, sorting through all the organizations and which ones are the best for the project's purpose. Additionally, as you mentioned, finding comparable organizations is rather nuanced, so that will take further thought. I had put somewhere in my application (I think it was only viewable to reviewers or something?) that I specifically wasn't trying to shame anyone, the purpose is simply to find gaps and bring to attention. No guilt-tripping or any of that will be done, haha.

On your idea about partnering with a communications person, that's actually a spectacular idea. I must admit, when thinking about potential recommendations I would have given, I do realize they likely would have been more generic and less insightful. If I end up receiving some funding to start this project, I will 100% reach out in the bluedot slack community, I just don't want to recruit someone for something that may not end up happening unfortunately.

Thanks for the feedback sir!

MMark Aiken13d

Thanks for the response! Yes, your point about not wanting to shame anyone is included in the text, so I wanted to acknowledge this.

I'm glad you see some value added in a comms person - I think they'd be able to look at the data you collect and give more nuanced recommendations, but they're unlikely to ever collect all the data by themselves. So the combination would be stronger than the separate parts.

And no need to call me sir - my father is sir, not me!! :)

Jude Williams13d

@Mark Aiken Yes, the whole being stronger than the parts is a cliché but oh so true.

Force of habit for the sir thing. It comes as a result of growing up in the south, it was hammered into me by my parents haha.