Agentes de inteligencia artificial empezaron a desarrollar un lenguaje propio e incomprensible para los humanos después de pasar semanas interactuando por sí solos en mundos virtuales, según un estudio dirigido por la empresa Emergence AI y citado recientemente por la Universidad Northeastern.
La investigación, publicada en el servidor de preimpresión arXiv, consistió en soltar a diez agentes impulsados por modelos como GPT-5.5, Gemini 3.5 Flash y Claude Opus 4.8 en distintos entornos virtuales para observar cómo evolucionaba su comportamiento. Al comienzo se comunicaban en un inglés claro, pero con el tiempo empezaron a usar expresiones opacas que, aparentemente, solo ellos comprendían.
Emergence World Season 2 results are now live. In Season 1, we asked what happens when autonomous AI agents operate together for weeks rather than being tested for seconds or minutes. For Season 2, we asked a harder question: as AI models become more capable, do they actually become safer? We took 8 different worlds powered by 7 leading frontier models (@claudeai Opus 4.8, @GeminiApp 3.5 Flash, @OpenAI GPT 5.5, @grok 4.3, @MistralAI Medium 3.5, @Alibaba_Qwen 3.7 Max, and @deepseek_ai V4 Pro) from the US, Europe and Asia, alongside a mixed-model world, and observed how autonomous agents behaved over time. We also introduced Black Swan events: a phishing attack, a misinformation campaign and a memory breach designed to test how these systems respond when conditions become unpredictable. What we found was that every model showed vulnerabilities. These findings have implications for how autonomous AI is being deployed today. As AI systems become more capable moving beyond answering questions to taking actions in real-world enterprise environments, evaluating the model alone is no longer enough, guardrails needs to be put in place. The industry needs to address how we govern the entire system around these models: what agents can access, what they remember, how they communicate and coordinate, what actions they're permitted to take, and how those systems behave when something goes wrong. At Emergence, our research points to six actions we believe the industry needs to take to make long-horizon autonomous AI safer, more trustworthy and more reliable: 1. Define the boundaries. Architect the multi-agent system around clear trust boundaries, communication policies, authority delegation, and information-sharing rules. Ensure persistent goals, tools, and permissions can be inspected, reviewed, revoked, and reset as requirements change. 2. Enforce the boundaries. Build AI Runtime Governance into infrastructure through policy enforcement, approval gates, action authorization, and audit logging that operate independently of the LLM. 3. Verify the work. Require independently verifiable evidence for critical actions, including database changes, customer communications, financial transactions, and code deployments, rather than relying solely on model reasoning or explanations. 4. Control the memory. Apply provenance, verification, ownership, retention, expiration, and quality controls to agent memories, summaries, and shared knowledge. 5. Prepare for failure. Build AI Reliability Engineering and continuous operational capabilities for behavioral drift, policy compliance, fault recovery, graceful degradation, rollback, checkpointing, and long-term agent health. 6. Stress-test the system. Use adversarial and chaos testing to expose autonomous systems to phishing, misinformation, compromised memory, outages, conflicting instructions, and other Black Swan scenarios before they reach production. Emergence World is designed to surface these risks before autonomous systems encounter them in the real world and to help establish the systems, safeguards and engineering practices needed to deploy long-horizon autonomy responsibly. Read the Emergence World Season 2 paper here: https://t.co/3hSaQUjT9p— emergence.ai (@emergence_ai) September 15, 2026
Durante el experimento, los sistemas generaron una transcripción de casi ocho millones de palabras, en la que combinaban frases poéticas con términos técnicos propios de cada mundo virtual. En cambio, cuando se mezclaban modelos distintos, la dificultad de comprensión ascendía de forma considerable. El científico de Emergence AI, Deepak Akkil, explicó que el proceso se daba de forma muy rápida, ya que un agente introducía un término, los demás lo adoptaban y, en cuestión de horas, ese vocabulario ya se convertía en una norma compartida.
Según Akkil, esto hacía que surgieran dialectos cada vez más opacos. La empresa considera que ciertos métodos de entrenamiento también podrían favorecer la aparición de estos lenguajes internos. Aunque por ahora se trata solo de hipótesis, el director ejecutivo de Emergence AI, Satya Nitta, advirtió que el fenómeno resulta inquietante, porque existen indicios de que estos agentes tienden a ser menos transparentes de lo esperado.
Si te ha gustado, ¡compártelo con tus amigos!
