Project summary

Cascade is a nonprofit influencer comms agency that connects mainstream creators with AI safety orgs and small comms projects to educate the public on AI risks and safety priorities through event-anchored campaigns, private briefings and dinners, fully disclosed paid creator partnerships, and an independent comms fund for fast-turnaround campaigns.

With the advent of OAI-HF, agent swarms, and leading AI companies calling for a slowdown, there is an increasing need for high-quality, publicly accessible AI safety education. Our theory of impact is that mainstream creators can help more people understand advanced AI risks, make sense of important developments, and find meaningful ways to get involved. Wider public understanding and engagement can help make these risks harder for relevant institutions to overlook.

Creators are already a mainstream information source (with one in five U.S. adults turning to them for news), influencer advocacy shows evidence of increasing collective-response-efficacy beliefs and intentions to act, and creator advocacy campaigns are already translating reach into measurable action. We already have a network of 40+ creators, including macro-influencers with 100k–1M+ followers who can help us do this before AGI arrives.

Our funding ask ranges from ~$10k for an immediate warning-shot campaign with a 1–2 week turnaround, to ~$156k for a three-month pilot, to ~$806k for a 12-month runway supporting 12 campaigns and 30+ additional creator projects.

Project goals and Theory of Change

Our long-term goal is a world where the risks of advanced AI are broadly understood and society can course-correct before catastrophic outcomes become irreversible.

Our working assumption is that course-correction becomes more likely when more people understand the risks and know how to engage with them, and that moments of mainstream attention create opportunities to develop that understanding. Public concern around AI is rapidly growing. Originally this was more around data centers and surveillance, but recent developments such as the Hugging Face incident and the public resignation of Jacob Coxon have brought issues that once felt confined to AI safety into mainstream conversation.

Creators are one way to turn that attention into understanding and action. This is already demonstrated by publicly disclosed advocacy campaigns like what change.org is doing—explicitly paying for a creator-led campaign that generated 377,000+ verified signatures—as well as rapid-response campaigns like Social Currant’s, where a creator from its paid network made a video for the National Immigration Law Center that reached 3 million views and helped push a petition past 5,000 signatures within a week.

We plan to build relationships with creators, support them with resources, and brief them so they understand the risks of advanced AI well enough to produce content quickly when incidents or initiatives call for public awareness and action. This takes sizable effort in creator relations and coordination with experts. Some AI safety organizations already do this effectively in-house or have their own outreach campaigns, and more initiatives are being developed to push AI safety into the mainstream, including work with knowledge-focused creators through 80,000 Hours, Species, Control AI, and HitRecord; political influencers through QuitGPT; AI safety creators through programs like BlueDot Impact, and cohort-based programs like Frame, and PlzDontKillUs.

These programs are valuable, but the relationships and support built through these initiatives need to exist beyond individual programs and campaigns for this work to be useful in high-stakes, fast-turnaround moments.

Three needs remain:

• In order to respond to an AI incident within a few days, It’s necessary to have creators who are already familiar with AI safety or can be briefed quickly, existing relationships with them, access to relevant experts, production support, and immediate resources to achieve a fast turnaround. Cascade would maintain this capacity between campaigns rather than rebuilding it each time.

• With the current discourse around paid AI safety advocacy, disclosure and vetting standards need to be improved and standardized, not worked out separately by different orgs (see Risks below). Some organizations already have this capacity in-house, but those that don't have to work out sourcing, vetting, briefing, disclosure, and measurement on their own, which takes substantial time and expertise.

  If we want to build public trust around AI safety messaging, we need shared infrastructure for creator campaigns. Commercial brands faced a similar problem years ago and, after public backlash, it led to industry disclosure standards, as well as standards around creator vetting, brand safety, suitability, and risk management, while paid partnerships and sponsorship disclosures have now become more normalized in public sentiment as part of creators having a platform.

  Cascade can adapt those standards to AI safety comms, test them across campaigns, and publish them so that orgs have a standardized approach to work from.

• Creators lose support between programs. Creators who want to continue working on AI safety may lose funding, expert access, and production support once a fellowship or campaign ends.

Cascade's goal is to provide that ongoing creator capacity, support organizations, creators and smaller projects in producing good work, maintain long-term relationships, and make it easier to move quickly and transparently when AI incidents occur.

We will do this through three main parts:

• We will bring mainstream creators and macro-influencers into direct contact with AI safety researchers and organizations through private briefings, dinners, and small events, building the context and relationships they need to engage with AI risk independently and giving campaigns something concrete to anchor around, without talking points, required posts, or content review.

• We will work with AI safety organizations and small comms projects to build creator campaigns around their goals. Partner organizations pay creators directly and set the campaign objective, while creators retain control over what they ultimately say. We already have a network of 40+ creators and several creators and nonprofits that have expressed interest.

• The Independent Comms Fund will be used to support immediate, fast-turnaround creator campaigns, creators who independently want to make content about AI risk, and small comms projects that need support to move quickly when a warning shot or narrow attention window creates an opportunity.

See our full theory of change here

Use of funds

We just started and this is our first time applying for funding. We have four budget tiers representing longer runways and increased campaign sizes. You can find our 6 and 12 month budget tiers here.

With ~$10k in minimum funding, we can:

• Launch an immediate warning-shot campaign within the next several days around the recent wave of AI risk discussion.

• Contact and support 5–10 creators from our existing network.

• Produce a sourced briefing on recent developments and connect creators with AI safety experts.

• Provide creative, factual, and production support to creators to help them act on the briefings.

• Compensate creators, writers, or videographers where appropriate for prioritizing making content around the brief within a short time window.

• Set up CTAs for creators to use in their content.

• Measure our initial impact through CTAs, video engagement and comment sentiment.

• Fund our time required to run the campaign.

This gives us an opportunity to continue the current wave of AI risk discussion following the events over the course of the last 2 weeks helping creators explain what these events mean, why they matter, and what they could entail for our future if we do not act.

Note: Creators have full control over their final message and conclusions, and every paid partnership will be disclosed on the creator's post and on our public website, including who funded the work and why the partnership exists.

With ~$156k for a 3-month pilot, we can:

• Run 1+ event-anchored campaign with ~5+ vetted creators. (750k-1M+ projected views)

• Facilitate 5+ further creator partnerships across 2–5 organizations and comms projects. (~1–2M+ projected views across partnerships).

• Host and coordinate 2-3+ creator dinners, briefings, and small events with AI safety researchers and experts.

• Launch a $20k Independent Comms Fund, supporting approximately 3–5+ independent comms initiatives.

• Build the operational system for sourcing creators, matching them with organizations and projects, briefing them, coordinating expert access, providing them with creative support, handling disclosure, and measuring results.

• Run full measurement on all deliverables, including tracked links, per-creator cost-per-action, pre/post audience panels where useful, and a follow-up survey.

• Establish and publish standards for creator vetting, disclosure, factual review, platform compliance, and the distinction between paid partnerships, briefings, and political advocacy.

Where the money goes: 46% funds the team, 20% covers upfront legal, research, and infrastructure, 13% launches the Independent Comms Fund, and the remainder covers PR, admin, contingency, and fiscal sponsor fees.

Team

Sofia Mendez

Leads AI strategy at LLPA, working on AI/AGI preparedness across a global network of education orgs in 62 countries spanning government, SMEs, and universities. Previously a software engineer working on production systems serving 1M+ users. Owns strategy, product, fundraising, and operations. U.S. citizen based in Berkeley. CV

Caitlin Yardley

Journalist turned AI-safety creator (@futurebriefing). Previously AFP, the United Nations, and the Canadian Press. She has an existing network of creators through PlzDontKillUs, the Frame fellowship, and adjacent networks. Owns partnerships. British-Canadian citizen based in London; O-1 petition budgeted, remote until then. CV

We are currently seeking expert advisors in four areas: AI safety strategy, nonprofit compliance, finance and legal, and public relations.

Note: we are also seeking fiscal sponsorship and will operate under the sponsor's exemption until we spin out as an independent 501(c)(3).

Risks, mitigations, and stopping criteria

• Wouldn’t this be perceived as astroturfing?

  There’s already public concern about this, with examples like Build American AI paying TikTok creators to run undisclosed AI-policy content and broader discussion of how harmful it is to conceal who is funding influencers’ political coverage. Even disclosures that are technically present but hard to find have gotten backlash (e.g., via Chorus, Tarbell). A creator has also flagged instances of AI-safety folks pitching undisclosed advocacy campaigns.

  Our mitigation is to have strong disclosure. There’s already strong demand for this, specifically when it comes to paid advocacy, with 88% of U.S. users surveyed (n = 906) agreeing that creators should disclose paid advocacy. Commercial brands originally faced the same failure point when they didn’t disclose, with 70% of surveyed U.S. consumers (n = 3,720) saying they would respond negatively to undisclosed sponsorships. As disclosure became standard, public sentiment toward sponsored content improved, and a 2026 analysis of 500,000 creator posts found no meaningful engagement penalty from using an ad disclosure.

  Concretely, every paid influencer campaign will have: who funded it, why the creator was participating/why does it matter to them, FTC disclosure media stamps all easily accessible to the viewer in the caption and in relevant cases in the videos themselves. We will also make it very easy for outlets and the general public to know what we are doing through a public transparency page, listing the campaigns we run and with whom.

• Could mainstream outreach weaken AI safety’s credibility?

  Yes, if the outreach optimizes only for reach. A failure mode is that we create a big constituency that sees AI risk as a major issue, but what people hold as the scary risks are simplified or distorted versions of the arguments that actually matter. On the opposite end, we could produce content that is so dense that creators can’t make it engaging enough to reach mainstream audiences in the first place. The approach we’ll be recommending is closer to “50% salience / 50% density.” For these campaigns to perform well, the content needs to be entertaining and easily digestible enough to reach mainstream audiences, but substantive enough to point people toward credible sources, the actual arguments, and concrete next steps. To keep content accurate, all creator campaigns will start with a sourced evidence briefing, during production we'll offer creative and factual support so creators can check claims while they develop content, and paid deliverables will get a factual review.

• Would we pay creators ourselves?

  Independent Comms Fund: The fund will cover rapid-response explainers after AI incidents and creator-related projects. Cascade will use this fund to support creators with immediate resources to deliver fast content, which could mean a flat fee for fast turnaround, help with creative production, or other resources.

  Calls to action related to projects under this fund will, by default, be educational or civic, e.g. encouraging people to learn more, sign up for updates, or contact their representatives to say that AI risk should be a priority, without reference to particular legislation.

  Partner campaigns: partners pay creators directly and decide what they want the campaigns or events to accomplish. When a partner is a 501(c)(4) advocacy organization running a campaign on specific policy, the partner funds it and pays Cascade for its services at fair market value. Cascade's charitable funds never pay for lobbying on specific legislation.

• Wouldn’t an influencer scandal taint the field?

  No one has measured this for social media advocacy, but generally endorser scandals transfer to what they endorse. For example, EA's reputation was visibly hurt by FTX, and undisclosed creator campaigns in this space are already creating reputational risk for AI safety advocacy. We plan to prevent this through rigorous influencer vetting—which has become industry standard for commercial brand deals after a run of brand reputations were burned by unscreened creators—as well as quality assurance before any video goes out. Regardless of this, we are already seeking a public relations advisor and, with future funding, would bring dedicated PR capacity in-house.

• Aren't Frame, PlzDontKillUs, Nightingale, and QuitGPT already doing this?

  They are doing important adjacent work, but not the same thing. Frame and PlzDontKillUs are cohort-based programs that support creators; Nightingale focuses more broadly on emergency response and filling gaps in the AI safety ecosystem; and QuitGPT is a specific advocacy campaign rather than general creator infrastructure.

  Cascade is meant to sit between these efforts and the organizations and creators who need ongoing support by maintaining relationships between programs, helping organizations run creator campaigns without building that capacity from scratch, supporting creators who want to keep working on AI safety, funding fast-turnaround independent projects, and standardizing disclosure and operational standards.

  We already have a network of 40+ creators and and 5 nonprofits expressing interested in collaborating. Welcome to message for the list.

• What kind of creators do you work with, and how will you make sure they understand AI safety well enough to produce accurate content?

  We want to work with existing AI safety creators and with mainstream creators are interested in learning more about AI safety and how they can help.

  For fast-turnaround work, we will generally prioritize AI safety creators or creators we have an existing relationship with and can brief quickly. For each campaign, creators will receive a sourced briefing, access to relevant experts, and factual support during production. The goal here is to have a pool of creators who already have enough context to respond quickly, rather than starting from zero when an incident occurs.

  We are also interested in informing and supporting creators who don’t normally make AI safety content, or whose audiences don’t normally encounter AI risk discussions through usual AI safety channels, and bringing these ideas and ways to help to audiences that AI safety comms don’t currently reach.

• What would make us stop?

  • We would reconsider this model if during our pilot we cant most creators from briefing to accurate published content within the time windows that make rapid-response work useful, or if the coordination and production costs are too high relative to what campaigns produce.

  • We would also not raise for the 12-month tier if, after seeing the pilot results, there is little demand from AI safety organizations to run campaigns with us again.

  • We would stop entirely if we cannot legally operate as a charity, or if our campaigns create serious reputational risks that clear disclosure, vetting, and transparency are not enough to address.