The "transcript" turned out to be an AI summary - and the court wasn't buying it
An Atlassian employee's bid to freeze performance reviews and enforce disability accommodations hit a wall in the Federal Court.
The worker, who represented himself, had filed Fair Work Act claims against the tech giant and two of its employees. He sought a string of urgent court orders: stop his employer from terminating him, halt its performance management process, compel it to implement reasonable accommodations for his neurodiversity, and excuse him from unmoderated one-on-one meetings with his direct supervisor.
The court, in a decision handed down on October 13, 2026, declined all four.
AI notes that didn't quite land
Central to the accommodations claim was a document the worker said recorded an agreement reached during a September 1, 2026, videoconference with another Atlassian employee. The worker said the meeting took place within Atlassian's corporate communications workspace and was transcribed by the company's "loom notetaker system."
Counsel for the respondents submitted the document was "not a transcript, but a summary." The court agreed - and went further, noting that "on the face of the document it appears that it (or some aspects of it) have been generated by the use of artificial intelligence."
Even taking the summary at face value, the court found the arrangements described were for a trial period of four to eight weeks and contingent on the worker providing further information and the company conducting further investigation. The court was "not persuaded that there were firm commitments that arrangements were to continue for an indefinite period."
It declined to make an order requiring Atlassian to implement the accommodations, citing the "imprecision and contingent nature of the arrangements discussed" and the likelihood that compliance would require ongoing court monitoring.
The paper trail started months earlier
The worker had argued the performance management process was retaliatory - triggered either by a letter of demand he sent on August 7, 2026, or by his refusal to accept a voluntary separation offer.
The evidence pointed the other way. The court found that performance concerns had been communicated to the worker from at least April 23, 2026 - well before the letter of demand. While the worker disputed that two of the relevant performance meetings took place, denied receiving certain supporting documents, and questioned their authenticity, the court noted there were other occasions where concerns had been raised before any legal complaint.
On the separation offer, counsel for the respondents pointed out that a closer reading of the documents showed the company was responding to a settlement proposal the worker himself had provided - not initiating a separation. Those same documents outlined a structured performance improvement plan, which the court found inconsistent with an inference that Atlassian's intention was to terminate.
The worker also argued that the performance concerns were really unaccommodated manifestations of his neurodiversity. The court accepted his evidence that he had been diagnosed with ADHD but noted there was no medical evidence before it about the condition's impact on his work performance, or how it might correlate with the specific concerns that had been raised.
No threat, no injunction
On termination, the court found simply no evidence it was being considered. Atlassian's senior director of employment law for the Asia-Pacific region deposed that neither she nor the worker's supervisor had discussed or considered termination. The supervisor separately deposed he had never made any decision to terminate and was not aware of any other manager at Atlassian having done so.
Atlassian gave an undertaking to provide one month's notice before implementing any termination or disciplinary action, and confirmed it would continue to manage the worker's employment through its standard performance management process. The court also preserved the worker's right to apply, on 48 hours' notice, for an order restraining his termination if circumstances changed.
Then there were the meetings
The worker had also sought an order excusing him from "unmoderated" one-on-one meetings with his supervisor - a request linked to a psychosocial hazard complaint he had lodged internally but which had not yet been investigated.
The court found two problems. First, the claim was not anchored in the worker's formal court filings or the final relief he was seeking - in short, it was unclear how the proposed order connected to the case before the court. Second, there was no medical evidence supporting a claim that unaccompanied one-on-one supervision would have a detrimental impact on the worker. Three of four medical certificates did not specify the relevant condition, and the fourth identified an unrelated one.
The court observed that Atlassian "might consider it prudent in the present circumstances for the meetings to be accompanied by" an employee relations staff member - but that the worker had not established a basis for the court to step in.
The matter has been referred for allocation to a docket judge. Interim confidentiality orders over the filed documents were extended.
For HR teams, the decision is a practical signal that AI-generated meeting summaries may not carry the evidentiary weight employees expect - and that performance management processes with a documented trail predating legal action are far harder to frame as retaliatory.
The proceeding remains before the Federal Court. No final determination has been made on the merits of the worker's Fair Work Act claims.
