Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Several image generators tested in September 2024 struggled to produce a consistent likeness of Kamala Harris—but that does not show that every AI model was unable to draw her, or reveal one definitive cause. The reported failures are better understood as a mix of unstable identity representation, possible demographic bias, uneven image labeling and model-specific safety rules. The episode was a reported set of tests and viral examples, not a controlled benchmark of all image generators, and it does not establish how current versions perform.
What happened in 2024?
The episode drew attention after Elon Musk shared a Grok-generated image depicting Harris as a supposed “communist dictator.” The image’s political framing was inflammatory; its subject also did not convincingly resemble Harris. Online users then compared Grok’s results for Harris with images it generated of Donald Trump.
WIRED subsequently reported trying prompts in Grok and finding that Harris’s facial features, hairstyle and skin tone varied across outputs. Some images appeared more like another public figure, including Michelle Obama. The outlet also reported weak likenesses from an open-source Stable Diffusion model. These observations describe the tested systems and prompts; they are not a standardized comparison across every model, checkpoint or setting. (WIRED’s report; Futurism’s account.)
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchSeveral commercial tools—including ChatGPT’s image tools, Google Gemini and Midjourney—were reported to refuse requests to depict politicians at the time. That matters: a refusal is not an inaccurate portrait, but it is not proof the system could draw the person accurately either.
#1 Best Overall
- Rich Creative Painting Ideas: This creative drawing prompt card deck set comes with 120 creativity prompt cards and 40 dry erase PVC blank cards, offering ample quantity and creative ideas for individual play, classroom activities, or parties with friends and family. The prompt cards provide endless drawing inspiration to spark your imagination.
- Simple and Enjoyable: This art challenge card deck set uses different colors to make the activity clearer and simpler: pink cards represent the subject, green cards represent the action, and yellow cards represent the scene. Simply draw one card from each color at random, combine them, and let your imagination flow as you create a picture on the blank PVC card. The PVC dry-erase cards are reusable and erasable.
- Promote Creative Inspiration: This art inspiration ideas set offers endless combinations, providing a unique drawing challenge every time. The creative drawing idea cards sparks creativity, guides artistic expression through drawing exercises, enhances observation skills and techniques, builds artistic confidence, and encourages you to step outside your comfort zone to explore different themes.
- Easy to Carry: Measuring approximately 8.7 x 6.2 cm (3.4 x 2.4 inches), these painting ideas cards are compact, lightweight, and come in exquisite boxes for easy storage and portability—ideal for creative fun anytime, anywhere. Beautifully packaged, this drawing & painting game set also makes a thoughtful gift.
- Wide Applications: This art prompt deck cards set is suitable for all skill levels and ages, from beginners to experienced artists. Ideal for artists, drawing enthusiasts, students, teachers, and anyone looking to improve their drawing skills, expand their imagination, and overcome creative blocks. Say goodbye to blank canvases and unlock your creative potential.
Why might a model get a public figure wrong?
Text-to-image generators generally do not look up a verified portrait and copy it. They synthesize an image from patterns learned between words, visual features, people and contexts. A model might learn associations between “Kamala Harris” and a podium, suit or hairstyle without forming a stable representation of her face. If that identity signal is weak, the result can drift toward a generic person or a visually associated figure.
More photographs do not automatically mean better representation. Images must be included in a model’s training data and identified consistently enough for the system to connect the name to the face. Captions and metadata may use different titles—senator, vice president, candidate or attorney general—or omit her name entirely. Image quality, cropping, pose and how frequently examples appear can also matter.
WIRED cited a Getty Images search snapshot of about 63,295 results for Harris and 561,778 for Trump. That suggests a difference in the volume of images available through that particular search, but it is not a count of either model’s training data. Training datasets are not established by a stock-photo search, and the figure cannot prove what any system saw or learned. Freepik CEO Joaquin Cuenca Abela offered the view that Harris was comparatively new to image-generation systems and that correctly labeled examples take time to accumulate; that is an informed explanation, not a confirmed account of the models’ datasets.
Rank #2
- Deck of 50 cards, each with a unique drawing prompt designed to inspire creativity and artistic exploration
- Uncoated smooth index-card-like 80# cover paper allows easy drawing with pencils, pens, or markers directly on the blank area of the card
- 2.5 x 3.5 inch size, similar to standard playing cards, fits easily in a pocket or bag
- Each card has a unique hashtag for sharing on social media
- Encourages creative practice and helps build a personal collection of art
Could race or gender bias be part of it?
It is plausible, but the Harris-specific cause was not isolated. Irene Solaiman, Hugging Face’s head of policy, suggested that poorer recognition of darker skin tones and feminine facial features could affect image sorting and labeling upstream. If a vision system detects, ranks or captions some faces less reliably, that can affect what examples a generator learns from.
Bias can also arise in the generation stage. A model may lean toward visually common examples associated with words such as “politician,” “leader” or “president,” or reproduce stereotypes in its training data. OpenAI’s DALL·E 2 system card discussed representational bias and uneven performance; broader studies have found demographic inaccuracies in generated images. Those findings support concern about the general mechanism, but they do not prove which mechanism produced a particular Harris image. (Research on demographic inaccuracies; research on racial bias in generated images.)
That is why neither “the model was deliberately programmed to make Harris look bad” nor “it was just random, with no possible connection to bias” follows from the available evidence. The observed instability is compatible with data-quality and demographic-bias problems; it does not establish deliberate intent, a specific discriminatory rule or a universal inability to render Harris.
Rank #3
- THERAPEUTIC WRITING CARD DECK — 100 DOUBLE-SIDED CARDS · 100 photographic image prompts + 100 text journal prompts, one on each side. 300gsm high-gloss. 11×11cm. Arrives in a hard cardboard keepsake box. Reusable — write in your own blank journal.
- GUIDED JOURNAL CARDS FOR WOMEN, TEENS & GIRLS · Morning intention-setting, evening self-reflection, or a daily self-care ritual during seasons of change and personal growth. Fits a purse, therapy bag, or nightstand — a gentle grounding ritual anywhere.
- IMAGE PROMPT JOURNAL + EXPRESSIVE WRITING PROMPTS FOR WOMEN · Photographic visuals spark honest reflection for art journaling, mindfulness, anxiety relief, and trauma work. Pair with a blank journal for deeper self-exploration. Images created using AI.
- SELF-CARE GIFT & THERAPIST GIFT — NOT A JOURNAL BOOK OR ORACLE DECK · Reusable, screen-free, and tactile. Meaningful for daughters, sisters, nieces, and friends — and a trusted therapy office tool for counsellors and facilitators.
- HEARTSCRIPT BY YOCHEVED ROTTENBERG — IFS PRACTITIONER & CERTIFIED JOURNAL FACILITATOR · Trauma-appropriate. Trusted by therapists in therapy offices and workshops. Includes instruction booklet. Use independently or alongside Write Your Way Home programs.
Why did Trump seem easier to draw?
Trump’s decades-long media presence may mean more images, more consistent labels or more repeated visual associations in some training collections. His distinctive, familiar appearance could also make a rough likeness easier to recognize. But those are plausible explanations, not a demonstrated causal result. Online sharing adds another complication: striking or recognizable outputs are more likely to circulate than ordinary failures, so viral comparisons can exaggerate a difference.
Free tools Windows power users keep installed
One-click scans. No signup required.
A fair test would need matched public figures and prompts, multiple generations and model versions, documented settings, and independent raters judging likeness. Without that, the comparison is an observed asymmetry—not proof that one person was favored, or that fame alone explains the result.
Why other services appeared to behave differently
Some competing systems did not produce a bad Harris portrait because they declined to generate politician images. OpenAI’s published DALL·E 2 safety work described measures to limit realistic likenesses of public figures. Google separately acknowledged that Gemini’s image-generation safeguards and demographic tuning had produced inappropriate results; in February 2024, it temporarily paused people-image generation while addressing that issue. That was a different incident, not evidence about the cause of Grok’s Harris outputs. (Google’s explanation.)
Rank #4
- 60 hand-drawn cards - Original pen-and-ink illustrations exploring shadow, pattern, and perception.
- 132-page companion booklet - Brief, open-ended reflections for each card.
- Made as a creative and reflective tool - Useful for journaling, self-directed spreads, or any practice that benefits from symbolic prompts.
- Independent artwork - Entirely hand-drawn with no AI or digital rendering beyond basic contrast adjustments for print clarity.
- Printed for durability - Cards use 330gsm stock with a protective gloss finish.
Safety policies differ by product and change over time. OpenAI’s current published political-campaigning restrictions are not a record of how every service behaved in September 2024. A model’s refusal should therefore be recorded separately from its likeness quality. “Would not generate” and “generated inaccurately” are different outcomes.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What a stronger comparison would require
To determine whether a model has a repeatable likeness problem—and what might cause it—an evaluation should identify the exact model, checkpoint, interface and test date. It should use fixed prompts and settings, disclose seeds where supported, and run multiple generations rather than relying on one striking result. It should compare Harris with public figures matched for factors such as gender, race, prominence and image coverage.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Independent raters could score identity likeness separately from skin tone, age, hairstyle and scene accuracy. Researchers should report refusals as their own result, preserve prompts and outputs, and replicate the test on more than one version. That kind of comparison could help separate prompt sensitivity, randomness, weak identity representation and demographic disparities. The 2024 reports are useful evidence of a problem worth investigating, but they do not supply that controlled causal test.
Why an obviously bad image can still matter
A poor likeness is not necessarily harmless. An image can attach a false costume, symbol or political identity to a real person even when the face is wrong. A viewer may see it outside the original prompt and context, or accept its message because a prominent account shared it. A synthetic image can also add to the broader difficulty of distinguishing authentic material from fabricated content.
That risk is different from whether an image is photorealistic. A strange-looking image may still function as satire, harassment or political propaganda when paired with a forceful caption. OECD.AI cataloged the episode in its incident database as relevant to misinformation and representation; the entry provides context, not proof of the technical cause.
The reported behavior belongs to September 2024. The available reporting does not establish that Grok, Stable Diffusion, Gemini, Midjourney or OpenAI’s current systems behave the same way. A present-day test would need to identify the exact model and version, date, region, prompt and settings. The incident is best read as a warning about unstable representations and uneven safeguards—not a universal 2026 benchmark or evidence of an anti-Harris switch.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
