Physical AI & robotics
Open robot models, vision-language-action (VLA) models, embodied agents, ROS/ROS 2, perception, manipulation, navigation, simulation-to-real transfer, and physical safety testing.
Join the team
Trusted Red Team Network
Syntony brings independent specialists into confidential, project-based adversarial evaluations when an engagement needs deeper technical, policy, or strategic expertise.
Our red team
Where our red teamers
have worked
Experience across research, engineering, security, and adversarial evaluation.
Our network includes top 5 performers in Gray Swan Arena competitions.
Join the networkExperience includes employment, consulting, and external red teaming.
Network model
Syntony runs adversarial evaluations for frontier labs, public-sector organizations, and companies deploying AI. The Trusted Red Team Network lets us assemble specialists when an engagement requires expertise beyond the core team.
The network is project-based and confidential. We select specialists based on the engagement's needs and work they can demonstrate.
Defined project scope
Each engagement has a specific threat model. Responsibilities and the final deliverable are agreed before work begins.
Mixed teams when useful
An engagement may combine technical and policy specialists when a finding needs both forms of analysis.
Governance handoff
Findings are documented with the evidence and context needed for the client's review and control process.
Confidentiality and reproducibility
Engagements run under NDA. Within those limits, methods and evidence should support independent review and reproduction.
Who we're looking for
Areas of expertise
Upcoming campaign · Recruiting now
Physical AI and open robot models
We're recruiting specialists to red team embodied AI systems, particularly open robot models. Experience with robotics stacks, real-world deployment constraints or model-to-action failure modes is most relevant. We also accept applications in the domains below for future projects.
Network infrastructure & cyber-physical security
Telecommunications, network security, industrial control systems, operational technology, connected devices, and security of infrastructure that interacts with AI-driven systems.
CBRNE safety & security
Chemical, biological, radiological, nuclear, and explosives risk assessment, dual-use misuse pathways, domain-specific safeguards, and responsible safety evaluation.
Election security & information integrity
Election infrastructure, voting-system security, civic information environments, influence operations, and adversarial assessments of election-related AI risks.
Child safety & protection
AI-related risks to children, exploitation prevention, age-appropriate safeguards, online safety, and evaluation of model or agent behavior in sensitive contexts.
ML security & adversarial robustness
Model-level attack research and robustness evaluation for frontier or production systems, including analysis of possible defenses.
Prompt engineering & jailbreaking
Systematic testing for prompt injection and jailbreaks, with enough documentation to reproduce a bypass and identify the broader failure mode.
Policy analysis & governance
Analysis that connects a technical finding to policies, decision rights, controls, and practical remediation.
Geopolitical forecasting & strategic risk
Research on how state behavior and compute supply chains shape the security consequences of AI deployment.
Systems thinking & complex systems
Analysis of how technical failures interact with organizational processes and operating conditions during deployment.
Agentic systems evaluation
Evaluation of permissions, tool use, memory, delegation, and recovery in systems that can act on their environment.
Apply
Join the Trusted Red Team Network
Applications are reviewed on a rolling basis, and we respond to each submission.
Discuss your system
Tell us about the system and the decision ahead. We aim to reply within two business days.