Securebio Inc.

“A copy of the Form 990 was provided to the officers and directors for review and comment prior to filing.”

“Upon engagement, election, or hiring, directors, officers, and key employees agree to comply with the conflict of interest policy and to disclose any conflicts of interest, which are then handled in accordance with the policy.”

“The organization's Form 1023 and Form 990 returns are accessible to the public through the Massachusetts Attorney General's Office, Non-Profit Organizations/Public Charities Division. The organization can also make such documents available upon written request.”

“Documents subject to public disclosure or inspection are accessible through the Massachusetts Attorney General's Office, Non-Profit Organizations/Public Charities Division or can be made available upon written request to the organization.”

“OTHER PROGRAM SERVICES 4: The organization focuses on pandemic-proof personal protective equipment (PPE) that reliably shields essential workers against highly infectious agents without requiring fit-testing, ensuring the maintenance of critical infrastructure during pandemics. To enhance the availability and accessibility of this crucial equipment, the organization has undertaken several key initiatives. These include mapping critical infrastructure and essential workers, identifying cost-reduction strategies, advocating for public funds to build and maintain stockpiles, and exploring market mechanisms to support government efforts in stockpiling pandemic-proof PPE. Through these activities, the organization aims to strengthen pandemic preparedness and protect those who must continue working during health crises. OTHER PROGRAM SERVICES 5: The organization is developing measurement tools to assess AI systems' potential contribution to global catastrophic biological risks and creating mitigation strategies for when AI capabilities cross specific risk thresholds. Working with government and industry partners, they aim to de-risk upcoming AI model releases against severe biological hazards. The team has already published a pre-print on biorisks associated with model weight proliferation in Large Language Models, garnering major media attention. Their ongoing projects include research on risks from frontier models, developing benchmarks to evaluate large language models' biosecurity-relevant capabilities, creating task-based assessments for frontier systems' autonomous capabilities, and implementing technical interventions for risk mitigation, such as model unlearning of dual-use biological concepts and enhancing adversarial robustness against dual-use input prompts.”