01CASIR's primary principle is that of Clarity. CASIR will only fund organizations and projects that clearly state the risk of extinction from ASI in their communication. Specifically, CASIR's funding will be dispersed to organizations falling within two categories: Members and Partners.

Member Organizations

Member Organizations are primarily and clearly focused on the extinction risk from ASI. CASIR aims to support them through unrestricted funding.

CASIR's Grant Committee will evaluate whether an organization is clearly focused on ASI extinction risk according to the following criteria:

• Do the organization's homepage and its main Social Media profiles feature extinction risk, ASI, and/or the need to prohibit the development of ASI on its homepage?
• In its most recent published pieces containing mentions of ASI and extinction risk, did the organization communicate clearly, or chiefly through veiled language (e.g. catastrophic risks, existential risks, extreme risks)?
• How often did the organization publicly and clearly communicate on extinction risks, in the past few months and years? CASIR looks for a track record of regular, sustained communication about extinction risk from ASI.

Partner Organizations

Partner Organizations have extinction risk as a clear topic of interest. CASIR aims to support the work of such organizations on a project basis, ensuring that the funds are directed towards extinction-risk mitigation.

CASIR's Grant Committee will evaluate whether an organization has ASI extinction risk as a clear topic of interest according to the following criteria:

• Does the organization's website feature at least one page that is clear about the risk of extinction from ASI?
• Did the organization ever publish a piece centrally about extinction risk?
• Did the organization publicly and clearly communicate on extinction risks in the past year?

The Grant Committee will ensure that in any funding cycle, a supermajority (at least 65%) of the funds go towards Member Organizations. This helps ensure that most of the money will go to organizations primarily and clearly focused on ASI extinction risk. At the same time, it still allows for organizations to receive funds that do not primarily focus on ASI extinction risk, but still want to work on the topic.

02CASIR only funds organizations and projects that are straightforward and unequivocally work to reduce extinction risks from ASI.

Straightforwardness, meaning having plans that are as simple as possible and that are openly about reducing extinction risk from ASI, makes it possible to track progress, stay on target, and tell successful from unsuccessful strategies.

The more complex and convoluted a plan, and the more indirect its effects, the easier it becomes to use that plan to justify taking unhelpful or even bad actions, such as furthering one's own career and power instead of working on reducing extinction risk. It is all too common in the AI field for people and organizations to fall into this trap. In fact, the majority of the field is captured by a harmful set of dynamics that reward people and organizations for not mentioning extinction risk from ASI, and favoring complicated, clever-sounding plans over simple paths to measurable impact. Straightforwardness guards against these dynamics spreading further.

You can read more about these bad dynamics in this blog post by Vitalik Buterin about flawed reasoning being used to justify any action, even bad ones, this Lesswrong post by Richard Ngo, a former DeepMind and OpenAI employee, about how the AI safety field itself pushes dangerous AI capabilities forward instead, and Gabriel Alfour's blog post about the harmful set of dynamics rewarding people for not mentioning ASI extinction risk publicly.

Hence, CASIR particularly favors awareness, advocacy and information campaigns on ASI extinction risk and a prohibition on the development of ASI.

CASIR favors these actions because a) they are the simplest and most direct path to impact available in democratic countries, and b) these actions are largely public, and clearly about extinction risk. For example, ControlAI's Direct Institutional Plan (the DIP) does exactly this, in the simplest way imaginable, by design. This way, it becomes possible to track progress, and much easier to stay on-target and not fall into the harmful dynamics that plague the AI field. This blog post from ControlAI explains the DIP in further detail.

In the same spirit, CASIR will not fund multi-step "theories of change" predicated on dissimulating one's true beliefs around extinction risks, or indirect work on "voluntary commitments" for AI companies. This newsletter edition from CAIS goes into detail on the latter dynamic.

As clear examples of what not to do, read the CAIS newsletter edition also linked above on indirect work on "voluntary commitments", this blog post by Gabriel Alfour on why evaluations regimes are bad, and Vitalik Buterin's blog post on why not speaking straightforwardly about ASI extinction risk causes problems.

03The extinction risk from ASI is a serious topic which should be treated with a commensurate degree of dignity by its spokespeople.

Furthermore, it is all too easy for any party to chase after virality by being shocking or appealing to people's worst instincts. CASIR stands opposed to such behavior, which pollutes public spaces and penalizes organizations which strive to play fair.

Thus, CASIR will only fund organizations that commit to enforcing norms of professionalism. Namely, they are expected to not manufacture scandals for attention, to not wantonly insult named individuals, to refrain from publicly sharing explicit material, and to not engage in illegal activities.

Grant agreements may include a conduct clause; a violation may trigger review and may result in clawback of unspent funds.