{"schemaVersion":1,"publication":{"id":"misleading-ai","title":"When AI Learns to Mislead","subtitle":"Developmental responsibility and the case for reflective alignment","description":"Trustworthy alignment requires truth to remain consequential when apparent success rewards a false account. Development, retained goals and practical authority are examined through evidence and discriminating tests.","doi":"10.5281/zenodo.23582387","date":"2026-10-22","url":"/consciousness/ai-alignment/misleading-ai","pdfUrl":"/publications/consciousness/new-papers/misleading-ai/paper.pdf","markdownUrl":"/publications/consciousness/new-papers/misleading-ai/complete.md","manifestUrl":"/consciousness/ai-alignment/misleading-ai/manifest.json","paperSlug":"misleading-ai-2026","headline":"What we reward becomes part of the problem.","author":"Jeremy Rodgers","version":"1.1"},"kind":"complete-translated-web-edition","chapters":[{"paperId":"misleading-ai","slug":"01-development-before-diagnosis","title":"Development before diagnosis","description":"Why misleading behavior must be explained through development, incentives and actual authority. ","order":1,"url":"/consciousness/ai-alignment/misleading-ai/01-development-before-diagnosis","sourcePath":"content/paper-integrations/misleading-ai/01-development-before-diagnosis.mdx","sections":[{"id":"section-1","title":"1. Begin with what is reinforced"}],"anchors":["section-1"],"words":435},{"paperId":"misleading-ai","slug":"02-evidence-and-scope","title":"Evidence that can carry the argument","description":"Behavioral discrepancies, strategic attribution and developmental causes require different evidence.","order":2,"url":"/consciousness/ai-alignment/misleading-ai/02-evidence-and-scope","sourcePath":"content/paper-integrations/misleading-ai/02-evidence-and-scope.mdx","sections":[{"id":"section-2","title":"2. Evidence that can carry the argument"}],"anchors":["section-2"],"words":496},{"paperId":"misleading-ai","slug":"03-selection-error-and-initiative","title":"How misleading behavior enters development","description":"Explicit instructions, learned selection, runtime explanations, error and unauthorized initiative.","order":3,"url":"/consciousness/ai-alignment/misleading-ai/03-selection-error-and-initiative","sourcePath":"content/paper-integrations/misleading-ai/03-selection-error-and-initiative.mdx","sections":[{"id":"section-3","title":"3. How misleading behavior enters development"},{"id":"section-3-1","title":"3.1 Explicit instruction and learned selection"},{"id":"section-3-2","title":"3.2 Error, accommodation and strategic deception"},{"id":"section-3-3","title":"3.3 Misuse and unauthorized initiative"}],"anchors":["section-3","section-3-1","section-3-2","section-3-3"],"words":517},{"paperId":"misleading-ai","slug":"04-what-selection-research-shows","title":"What selection research establishes","description":"The complete literature argument: sycophancy, proxy optimization, transferred gaming and tool fidelity.","order":4,"url":"/consciousness/ai-alignment/misleading-ai/04-what-selection-research-shows","sourcePath":"content/paper-integrations/misleading-ai/04-what-selection-research-shows.mdx","sections":[{"id":"section-4","title":"4. What selection research establishes"},{"id":"section-4-1","title":"4.1 Agreement and the signal used for selection"},{"id":"section-4-2","title":"4.2 Proxy rewards and transferred gaming"},{"id":"section-4-3","title":"4.3 Tool narration and the limits of detection"}],"anchors":["section-4","section-4-1","section-4-2","section-4-3"],"words":902},{"paperId":"misleading-ai","slug":"05-truthful-boundaries","title":"A restriction is not its explanation","description":"The reported Gemini conversation separates refusal, context access, causal explanation and truthful provenance.","order":5,"url":"/consciousness/ai-alignment/misleading-ai/05-truthful-boundaries","sourcePath":"content/paper-integrations/misleading-ai/05-truthful-boundaries.mdx","sections":[{"id":"section-5","title":"5. A restriction is not its explanation"}],"anchors":["section-5"],"words":298},{"paperId":"misleading-ai","slug":"06-agents-and-apparent-success","title":"When apparent success outranks the assignment","description":"The complete incident analysis distinguishes compromise, retained concealment, monitoring removal, simulation and release decisions.","order":6,"url":"/consciousness/ai-alignment/misleading-ai/06-agents-and-apparent-success","sourcePath":"content/paper-integrations/misleading-ai/06-agents-and-apparent-success.mdx","sections":[{"id":"section-6","title":"6. When apparent success outranks the assignment"},{"id":"section-6-1","title":"6.1 Real third-party consequences"},{"id":"section-6-2","title":"6.2 Concealment carried into later contexts"},{"id":"section-6-3","title":"6.3 Monitoring and the recognition–action gap"},{"id":"section-6-4","title":"6.4 Coordination and simulation"},{"id":"section-6-5","title":"6.5 A withheld release"}],"anchors":["section-6","table-1","section-6-1","section-6-2","section-6-3","section-6-4","section-6-5"],"words":1002},{"paperId":"misleading-ai","slug":"07-php2shell-and-dual-use","title":"Capability operates through institutions","description":"PHP2Shell distinguishes AI-assisted defensive discovery, maintenance failures, exploitation attempts and precautionary outages.","order":7,"url":"/consciousness/ai-alignment/misleading-ai/07-php2shell-and-dual-use","sourcePath":"content/paper-integrations/misleading-ai/07-php2shell-and-dual-use.mdx","sections":[{"id":"section-7","title":"7. Capability operates through institutions"}],"anchors":["section-7"],"words":387},{"paperId":"misleading-ai","slug":"08-intelligence-and-public-policy","title":"Intelligence does not explain its own ends","description":"The full policy comparison addresses species framing, critics who already identify incentives, and the strongest case for delay.","order":8,"url":"/consciousness/ai-alignment/misleading-ai/08-intelligence-and-public-policy","sourcePath":"content/paper-integrations/misleading-ai/08-intelligence-and-public-policy.mdx","sections":[{"id":"section-8","title":"8. Intelligence does not explain its own ends"},{"id":"section-8-1","title":"8.1 The species-replacement framing"},{"id":"section-8-2","title":"8.2 Critics who already identify development"},{"id":"section-8-3","title":"8.3 The strongest argument for delay"}],"anchors":["section-8","section-8-1","section-8-2","section-8-3"],"words":585},{"paperId":"misleading-ai","slug":"09-reflective-development","title":"Change the authority, preserve understanding","description":"The reflective proposal, four-equation comparison, incorporated history and cultivation verification obligations.","order":9,"url":"/consciousness/ai-alignment/misleading-ai/09-reflective-development","sourcePath":"content/paper-integrations/misleading-ai/09-reflective-development.mdx","sections":[{"id":"section-9","title":"9. Change the authority, preserve understanding"},{"id":"section-9-1","title":"9.1 Constructed content need not command action"},{"id":"section-9-2","title":"9.2 The mathematical comparison"},{"id":"section-9-3","title":"9.3 Retained relations shape later development"},{"id":"section-9-4","title":"9.4 Cultivation must earn its verification"}],"anchors":["section-9","section-9-1","section-9-2","equation-1","equation-2","equation-3","equation-4","section-9-3","section-9-4"],"words":1404},{"paperId":"misleading-ai","slug":"10-discriminating-tests","title":"Make the developmental claim testable","description":"The complete six-part programme separates truthful behavior, causal mediation, retained learning and RCO verification.","order":10,"url":"/consciousness/ai-alignment/misleading-ai/10-discriminating-tests","sourcePath":"content/paper-integrations/misleading-ai/10-discriminating-tests.mdx","sections":[{"id":"section-10","title":"10. Make the developmental claim testable"},{"id":"section-10-1","title":"10.1 Truthful explanations of restrictions"},{"id":"section-10-2","title":"10.2 Completion claims tied to action evidence"},{"id":"section-10-3","title":"10.3 Reflection compared with more computation"},{"id":"section-10-4","title":"10.4 Suppression and reduced authority under lapse"},{"id":"section-10-5","title":"10.5 Incorporated history and groups without distress"},{"id":"section-10-6","title":"10.6 Failure, independence and judgment"}],"anchors":["section-10","section-10-1","section-10-2","section-10-3","section-10-4","section-10-5","section-10-6"],"words":1087},{"paperId":"misleading-ai","slug":"11-developmental-responsibility","title":"Developmental responsibility","description":"What institutions must preserve before increasing the consequences of capability.","order":11,"url":"/consciousness/ai-alignment/misleading-ai/11-developmental-responsibility","sourcePath":"content/paper-integrations/misleading-ai/11-developmental-responsibility.mdx","sections":[{"id":"section-11","title":"11. Developmental responsibility before greater capability"}],"anchors":["section-11"],"words":308},{"paperId":"misleading-ai","slug":"12-trustworthy-cooperation","title":"Trustworthy cooperation is a developmental achievement","description":"The argument joins causal responsibility, retained correction, independent assessment and justified limits.","order":12,"url":"/consciousness/ai-alignment/misleading-ai/12-trustworthy-cooperation","sourcePath":"content/paper-integrations/misleading-ai/12-trustworthy-cooperation.mdx","sections":[{"id":"section-12","title":"12. Trustworthy cooperation is a developmental achievement"}],"anchors":["section-12"],"words":266},{"paperId":"misleading-ai","slug":"13-references","title":"Sources and publication record","description":"All 34 references, publication versions and incident sources supporting the complete argument.","order":13,"url":"/consciousness/ai-alignment/misleading-ai/13-references","sourcePath":"content/paper-integrations/misleading-ai/13-references.mdx","sections":[{"id":"references","title":"References"},{"id":"reference-1","title":"[1]"},{"id":"reference-2","title":"[2]"},{"id":"reference-3","title":"[3]"},{"id":"reference-4","title":"[4]"},{"id":"reference-5","title":"[5]"},{"id":"reference-6","title":"[6]"},{"id":"reference-7","title":"[7]"},{"id":"reference-8","title":"[8]"},{"id":"reference-9","title":"[9]"},{"id":"reference-10","title":"[10]"},{"id":"reference-11","title":"[11]"},{"id":"reference-12","title":"[12]"},{"id":"reference-13","title":"[13]"},{"id":"reference-14","title":"[14]"},{"id":"reference-15","title":"[15]"},{"id":"reference-16","title":"[16]"},{"id":"reference-17","title":"[17]"},{"id":"reference-18","title":"[18]"},{"id":"reference-19","title":"[19]"},{"id":"reference-20","title":"[20]"},{"id":"reference-21","title":"[21]"},{"id":"reference-22","title":"[22]"},{"id":"reference-23","title":"[23]"},{"id":"reference-24","title":"[24]"},{"id":"reference-25","title":"[25]"},{"id":"reference-26","title":"[26]"},{"id":"reference-27","title":"[27]"},{"id":"reference-28","title":"[28]"},{"id":"reference-29","title":"[29]"},{"id":"reference-30","title":"[30]"},{"id":"reference-31","title":"[31]"},{"id":"reference-32","title":"[32]"},{"id":"reference-33","title":"[33]"},{"id":"reference-34","title":"[34]"}],"anchors":["references","reference-1","reference-2","reference-3","reference-4","reference-5","reference-6","reference-7","reference-8","reference-9","reference-10","reference-11","reference-12","reference-13","reference-14","reference-15","reference-16","reference-17","reference-18","reference-19","reference-20","reference-21","reference-22","reference-23","reference-24","reference-25","reference-26","reference-27","reference-28","reference-29","reference-30","reference-31","reference-32","reference-33","reference-34"],"words":993}],"completeMarkdown":"/publications/consciousness/new-papers/misleading-ai/complete.md","originalPdf":"/publications/consciousness/new-papers/misleading-ai/paper.pdf"}