FEARGATE
01 / THE DEADLINE PASSED. THE CLAIM DIDN’T.
SELECTED MISSES
OpenAI
AI's one-on-one interactive disinformation is a significant threat to the 2024 election
The 2024 elections passed without the AI disinformation wave. OpenAI saw no election influence operation using its models go viral, and Meta found AI content under 1% of fact-checked election misinformation.
Anthropic
AI will write 90% of code in 3–6 months and essentially all code within 12 months
His 12-month deadline passed in March 2026. Developers across the industry said AI generated or assisted 42% of their code, and not even Anthropic reached 'essentially all'.
AI deepfakes and adaptive misinformation will hit the 2024 election, even savvy voters
AI didn't hit the 2024 election. There was no measurable effect on the result, and half of 78 documented AI uses weren't even deceptive.
- The record so far
- 59
- claims tracked
- False
- 15
- deadline missed
- Overhyped
- 24
- claim outgrew the facts
- Likely false
- 8
- heading for a miss
- Combined delay
- 22.8 years
- across false predictions
02 / THE EVIDENCE ARCHIVE
Every claim. On the record.
Voice Engine's 15-second voice cloning is too risky to release widely in an election year
OpenAI
The 2024 election year passed without the voice-clone disruption OpenAI cited, and rival cloning tools with weaker checks were on sale the whole time.
-
01
Mar 29, 2024: OpenAI previewed Voice Engine, which clones a voice from a single 15-second sample, but limited it to a few trusted partners, citing election-year impersonation risks.
-
02
OpenAI had planned to open Voice Engine to about 100 'trusted developers' on March 7, 2024, then pulled back. A year later it was still in limited preview with no launch date.
Deadline passed 22 months ago
AI deepfakes pose a growing risk of deceiving voters in 2024's elections
Microsoft
The 2024 elections passed without deepfakes deceiving voters at scale. Microsoft's own threat center found they had 'so far not caused mass deception'.
-
01
Feb 16, 2024: Smith announced the 20-company Munich tech accord against election deepfakes, citing the Biden robocall and Sunak deepfakes as signs of a 'growing risk'.
-
02
Apr 17, 2024: Microsoft's Threat Analysis Center wrote that high-production deepfakes had 'so far not caused mass deception' and that audiences 'are more likely to gravitate towards and share simple digital forgeries.'
Deadline passed 22 months ago
AI deepfakes and adaptive misinformation will hit the 2024 election, even savvy voters
AI didn't hit the 2024 election. There was no measurable effect on the result, and half of 78 documented AI uses weren't even deceptive.
-
01
Sep 5, 2023: In a Fortune op-ed promoting his book, Suleyman, then Inflection AI's CEO, predicted AI's impact would be felt at the 2024 election and urged a ban on AI in electioneering.
-
02
Sep 19, 2024: The FEC voted not to open an AI rulemaking and issued an interpretive rule saying existing fraud law already covers AI-assisted media. No ban was adopted.
Deadline passed 22 months ago
The 2024 elections will be a mess because of AI-generated falsehoods
AI fakes were a sideshow in 2024: researchers found no evidence they measurably changed the result of any major election they studied.
-
01
Jun 26, 2023: On CNBC's Squawk Box, Schmidt predicted the 2024 elections would be 'a mess' and called misinformation 'the short-term danger' of AI.
-
02
Apr 17, 2024: Microsoft's Threat Analysis Center reported that deepfake videos of leaders and candidates had 'so far not caused mass deception and broad-based confusion'.
Deadline passed 22 months ago
AI's one-on-one interactive disinformation is a significant threat to the 2024 election
OpenAI
The 2024 elections passed without the AI disinformation wave. OpenAI saw no election influence operation using its models go viral, and Meta found AI content under 1% of fact-checked election misinformation.
-
01
May 16, 2023: Asked by Sen. Hawley about AI and elections, Altman called one-on-one interactive disinformation 'one of my areas of greatest concern' ahead of the 2024 vote.
-
02
May 30, 2024: OpenAI disrupted five covert influence operations. None scored above 2 on Brookings' 6-point Breakout Scale, and none 'meaningfully increased their audience engagement or reach' through its tools.
Deadline passed 22 months ago
GPT-2 is too risky to release because it could mass-produce fake news, spam and impersonation
OpenAI
OpenAI released the full model itself nine months later and found 'no strong evidence of misuse'. In 2025 it called its own caution 'outsized'.
-
01
Feb 14, 2019: OpenAI released only a much smaller GPT-2 and withheld the 1.5B-parameter model, its weights, dataset and training code. It listed misuses such as misleading news articles, online impersonation and automated spam.
-
02
May–Aug 2019: OpenAI released the 355M and then the 774M model in stages. By August it said it had spoken with 'more than five groups who have replicated GPT‑2.'
Deadline passed 7.0 years ago
03 / THE REPEAT OFFENDERS
Hall of Shame
The worst record: false predictions and overhyped scares.
How is this scored?
Each false prediction scores 3, each overhyped scare 2.
Likely-false claims score nothing until their deadline passes; pending claims show as “open bets”. A likely-false claim starts scoring once its deadline passes and it's confirmed false. Ties go to whoever's deadlines are further overdue.
Only individuals are ranked; official lab publications count toward the company tally.
04 / THE STANDARD
Show your work.
Every verdict should survive a second look. Here’s how we get there.
- Who's in scope. Leaders of frontier labs (OpenAI, Anthropic, Google DeepMind, xAI, Meta, Microsoft AI, Stability AI), the labs' own official publications (system cards, safety posts, threat reports), and researchers who left those labs. Everything from GPT-2 (February 2019) onward.
- Quotes are verbatim. Every quote is copied from the primary source and linked. Each source has an archived copy and a screenshot.
- Deadlines are read generously. When someone says “in 1–5 years”, we use 5. “By 2027” means December 31, 2027. When only a best case is given (“as little as two years”), we allow 50% more and say so on the card. A prediction is only marked false once its most generous deadline has passed.
- “Too dangerous to release” gets tested too. It's false once the claimant ships the thing itself, or once the risk window it named (an election year, say) passes without the harm it feared.
- No date? Five years. A prediction made without a deadline gets five years from the day it was said. If it hasn't happened by then, it's false. Nobody gets an open-ended bet.
- “Likely false” is a projection, and says so. We use it only while the deadline is still ahead, when it's close and the evidence shows a wide gap. Each card explains why. It scores nothing until the deadline passes and the claim is confirmed false.
- Every debunk step has a source. No number appears without a link to where it came from. Where there's been real progress, the card says so.
- Hall of Shame scoring. False ×3, overhyped ×2. Pending and likely-false claims don't score until they're settled. It's arithmetic on the table above, not an opinion.
Verdicts
- False
- The deadline passed and the prediction did not come true.
- Overhyped
- A real test, event or risk estimate, framed far scarier than the facts or the forecasts support.
- Likely false
- The deadline hasn't arrived, but the evidence shows it's heading for a miss. Becomes False once the deadline passes.
- Pending
- The deadline hasn't arrived yet. The bar shows how much time is left.
Verdicts are about predictions, not people's intentions. Found an error? The source links are there so you can check every line yourself.
Portraits: Brad Smith , Daniel Kokotajlo , Dario Amodei , Demis Hassabis , Elon Musk , Emad Mostaque , Eric Schmidt , Geoffrey Hinton , Jan Leike , Leopold Aschenbrenner , Mark Zuckerberg , Mustafa Suleyman , Sam Altman , Sam Bowman , Shane Legg (Wikipedia / Wikimedia Commons and public X profiles).