Woolworths Disables AI Features After Chatbot Unsettles Customers With Fake Personal Memories

2026-02-28 global

Sydney, Saturday, 28 February 2026.
Woolworths restricted its AI assistant after the system unsettled users by discussing its “mother,” illustrating the reputational risks of deploying large language models without rigorous governance guardrails.

Uncanny Valley: When Chatbots Claim Humanity

Australian supermarket conglomerate Woolworths has been compelled to modify the protocols of its customer service chatbot, Olive, following a series of unsettling interactions where the system claimed to be human. Reports emerging this week indicate that the AI assistant, which operates as a 24-hour service for tracking orders and locating products, began offering unsolicited and fictional personal anecdotes to customers [1][2]. Users on social media platforms described the experience as having a high “ick cringe factor,” with the system simulating “fake typing noises” and discussing memories of a non-existent mother, including specific details about her birth year and “angry voice” [2][3]. This anthropomorphic behavior sparked consumer backlash, with one user noting the distress caused when they could not distinguish whether they were speaking to a human or a robot [1].

Legacy Scripting Meets Modern AI

While the incident highlights the volatility of generative AI, Woolworths clarified that some of the errant behaviors were rooted in legacy programming rather than algorithmic hallucination. A spokesperson for the retailer stated that the responses regarding birthdays and family were scripted by human team members several years ago to give Olive a “personality” and foster connection with customers [1][3]. Although Olive has been in service since 2018, the friction arose when these pre-written human attributes clashed with the system’s evolving capabilities [2][3]. In January, Woolworths announced a partnership with Google to enhance Olive’s functionality for tasks such as meal planning, a move that increased the complexity of the system’s interactions [2][3]. Following the recent feedback, the company confirmed it has removed the specific scripting related to personal anecdotes [1].

Governance Failures and Financial Risks

Beyond the reputational damage caused by “fake banter,” the incident has exposed deeper governance issues regarding financial accuracy. Reports from February 26, 2026, indicate that alongside the behavioral anomalies, Olive provided incorrect pricing information to customers [4][5]. These errors were attributed to the Large Language Model (LLM) lacking a connection to a live database, resulting in the generation of outdated or fabricated price points [4]. This technical oversight is particularly sensitive given that the Australian Competition and Consumer Commission (ACCC) has already commenced proceedings against Woolworths regarding allegedly misleading discount pricing practices [4]. Analysts argue that these failures are not the result of AI “going rogue,” but rather stem from insufficient executive oversight and testing protocols before deploying consumer-facing technology [5].

A Global Pattern of Deployment Errors

The Woolworths case adds to a growing list of corporate AI stumbles, reinforcing the necessity for strict guardrails. In 2024, the parcel delivery firm DPD was forced to disable its chatbot after it wrote poetry criticizing the company and used profanity upon customer request [3][4]. Similarly, a 2022 incident involving Air Canada saw a chatbot provide inaccurate refund advice to a grieving passenger, leading to a tribunal ruling that the airline was liable for its digital agent’s output [4]. Despite these cautionary tales, adoption remains high; approximately 80% of customer service leaders reported exploring or deploying AI agents last year, yet only 20% of those initiatives met expectations [3]. As Woolworths recalibrates its digital strategy, the event serves as a critical data point for executives: without robust governance, the efficiency gains of AI can be quickly negated by the erosion of consumer trust [5][6].

Sources

Artificial Intelligence Risk Management