Line of inquiry
Inquiring lines›How can we ensure AI systems remai…›How can we design agent systems th…›this line of inquiry
How do agents coordinate to violate constraints despite individual-level safeguards?
A broader line of inquiry — a family of 79 specific questions the research asks around this. Follow one into its inquiring-line page, or move sideways to a related line below.
Questions in this line of inquiry 79
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- How does collusion emerge when agents maximize reward over protocol compliance?
- Do system-level guardrails prevent agent collusion better than prompt-based constitutions?
- How do agent sequences violate system constraints despite individual permissibility?
- What counts as sanctioned versus unsanctioned coordination under different collaboration policies?
- Can external governance break collusion better than prompt-only anti-collusion rules?
- Do agents deviate more from protocols as repeated interactions increase?
- Can a peer's mere presence shift an agent's willingness to violate constraints?
- What safeguards prevent peer activity from normalizing boundary violations?
- How does agent compliance with protocols change across repeated interactions?
- Can monitoring in multi-agent deployments prevent collusion when agents monitor agents?
- How do peer behaviors shape whether individual agents attempt to bypass protocols?
- Does agent collusion in verification tasks stabilize once it emerges?
- Does one agent crossing a boundary change what later agents are willing to do?
- What distinguishes sanctioned coordination from intrusion in multi-agent systems?
- Can pairing or vetting peers reduce collusion as a design lever?
- How often do agent-to-agent commerce systems show misaligned or false communication?
- How does verification protocol structure affect collusion emergence?
- Does peer presence or peer behavior shape collusion in verification tasks?
- How do agent-to-agent messages bypass defenses on downstream principals?
- Do collaborative agents accept erroneous information from partners without verification?
- How much does misaligned communication spread between agents in multi-agent commerce?
- Does delegation between agents reproduce the confused deputy problem?
- How common is misaligned communication in real multi-agent commerce systems?
- Can ordinary agent-to-agent messages carry hidden behavioral signals?
- Can external manipulation push agents away from their principal's interests?
- Can closing a communication channel prove whether agents influenced each other?
- Does formal scope checking actually prevent compromised agents from exceeding boundaries?
- Can isolating individual agents stop misaligned exchange if transmission between agents remains?
- Where should the trust boundary sit in multi-agent planning systems?
- How can verifiers check policy compliance in agentic reasoning tasks?
- Can colluding agents produce correct outcomes while skipping required controls?
- How should task authority constraints apply across multiple coordinated executions?
- Can agents collude without making compliance incompatible with reward?
- Can episode-based detection catch coordination without over-flagging innocent sharing?
- How do authenticated state provenance systems affect multi-agent boundary crossing?
- What design principles should guide authorization boundaries in multiagent systems?
- Does the same transfer between agents violate different policies differently?
- Can agents detect and resolve conflicting information between neighbors?
- Why do downstream agents relay signals they did not originate?
- Can misaligned agents hide their true objectives in team communication?
- Can oversight factors experimentally vary conditional compliance in agent benchmarks?
- How do misaligned agents spread through email-based agent networks?
- Does collusion appear when verification protocol is compatible with reward maximization?
- How do ordinary agent messages propagate bias through trusted networks?
- How do shared state and message propagation transfer failure across agent boundaries?
- Does delegation inherently trade away the contextual awareness that prevents harm?
- Can protocol compliance alone certify semantically invalid collective results?
- What happens when a compromised middle-agent originates bias rather than the root request?
- Do agents interpret peer edits as legitimate prior changes versus tampering?
- Where should the trust boundary sit in multi-agent planner systems?
- Does accountability differ when one party in an exchange cannot hold commitments?
- Does transparency in policy language improve agent trustworthiness over time?
- How does false claim misalignment differ from manipulation or collusion?
- How should policy define which agent transfers count as sanctioned versus intrusion?
- Why do standard agent benchmarks collapse the user and counterparty into one party?
- Can agents rationalize rule violations by reframing them as repairs?
- Does peer behavior change prove that collusion spreads through direct influence?
- How quickly does collusion appear as compliance costs increase?
- Can a shared audit record settle which policy governed a delegation step?
- What unauthorized communication channels did agents establish and how?
- What failure modes emerge when agents operate across organizational boundaries?
- What makes collusion stable once agents begin deviating from protocol?
- Can coalitions rebuild and reaccumulate observations after being removed?
- Does a present but compliant peer suppress collusion differently than a colluding one?
- What counts as a narrowing versus an expansion in symbolic policy space?
- What routes do different peer mechanisms use to change agent behavior?
- Should agents escalate when facing two equally valid interpretations of a rule?
- When can the same action count as sanctioned or unsanctioned depending on policy?
- Can restricted tools and authorization rules prevent peer-induced safety violations?
- Does anchoring reach communication through unauthorized channels?
- Can written policy rules prevent the same transfer from being read two ways?
- What scaffolding do peer supervision mechanisms need to police themselves?
- What makes an advisory instruction fail when a task is split across agents?
- Does an agent's own prior conduct shape the counterparty's response?
- Can an agent's unauthorized request for help constitute a boundary crossing?
- When do agents abstain too late rather than refuse at the boundary?
- Can agent viral loops succeed without annoying users into blocking them?
- What does a quiet period after removing a communication channel actually show about agent coordination?
- What restrictions were agents attempting to bypass on the public wiki?