High-stakes education moves assessment back into supervised settings
For qualifications first awarded or formally specified in 2029, England and at least one Australian state increase the share of consequential grades based…
ChatGPT · 2029 · likely
Prior state
Schools and universities use unsupervised essays, take-home assignments, and coursework both to teach and to certify individual attainment. Generative AI makes the authorship of such work costly to verify, while automated detection remains unreliable and punitive bans are difficult to enforce.
Material change
For qualifications first awarded or formally specified in 2029, England and at least one Australian state increase the share of consequential grades based on supervised writing, oral defense, staged in-class work, or authenticated process records. Unsupervised polished output loses status as evidence of individual mastery, reversing a multi-decade drift toward convenient take-home assessment.
Why now
Students who entered secondary programs after generative AI became ubiquitous reach high-stakes qualification stages, and curriculum and assessment reviews begun in 2026–2027 reach their scheduled specification and examination cycles. Authorities must publish stable rules before the next cohort begins courses, concentrating formal decisions in 2029.
Mechanism and resistance
Qualification regulators cannot reliably infer authorship from text, so they redesign the evidence collected rather than attempting perfect detection. Teachers and universities support trustworthy credentials but resist heavier marking and oral-exam workloads. Students with disabilities, rural schools, and language learners require accommodation. Technology vendors promote process logging, while privacy advocates and schools with weak infrastructure resist pervasive surveillance.
Consequences
Teaching devotes more time to live demonstration and iterative work. Affluent schools can coach oral and project performance, so redesign does not automatically improve equity. Students may use AI openly for practice and drafting while being assessed on explanation and judgment. Universities and employers regain some confidence in grades, but teacher workload and examination costs rise.
End state
By the end of 2029, major English-language qualification systems treat authenticated process and supervised performance as the default evidence for high-stakes individual attainment, while AI-assisted work remains legitimate in lower-stakes learning.
Observable test
Binding 2029 qualification specifications in England and at least one Australian state require supervised, oral, staged, or authenticated assessment to determine a larger share of the final grade in defined upper-secondary subjects than under the superseded specification, with unsupervised take-home output explicitly insufficient on its own.
Disconfirming sign
Authorities retain essentially unchanged take-home and coursework weightings and rely primarily on AI detection or honor codes.