Topic: #meta
-
Meta's TextSeal2 watermark detects 89% of AI text in researchers' tests
Meta researchers say TextSeal2, a new LLM watermarking technique, detected 89% of 200-token AI-generated outputs at a 0.1% false-positive rate, ahead of the original TextSeal's 71%. Meta hasn't announced a product deployment.
-
Meta and NYU's h1 matches or beats elaborate Lean workflows at lower cost, paper says
Meta FAIR and NYU researchers released h1, a minimal AI loop linking frontier LLMs to Lean 4. Their paper says it scored 670/672 on PutnamBench and spent $121 per proof versus $317 for AlphaProof Nexus.
-
Nvidia to put $3 billion into Meta's $95 billion AI fundraising, report says
Nvidia planned to invest $3 billion in Meta Platforms, Crypto Briefing reported, as Meta laid out a two-stage plan to raise roughly $95 billion for its artificial intelligence infrastructure buildout.
-
Meta plans up to $95 billion stock sale as AI spending outlook climbs
Meta Platforms said it plans to raise up to $95 billion by selling stock, its largest equity fundraising ever, as it lifts its 2026 capex outlook to as much as $145 billion for data centers, chips and AI.
-
Meta drops AI training language from its commercial terms after publisher backlash
Meta removed language from its Commercial Terms that appeared to let it scrape outside websites, including news sites, to train its AI, Crypto Briefing reported, after criticism from publishers.
-
Meta takes back $25 million from midterm push as ATEP super PAC returns $22M
Meta has pulled $25 million back from its midterm political push: $22 million returned by its ATEP super PAC and $3 million by a California advocacy group, according to Crypto Briefing.
-
Meta unveils Muse Wonder AI models, pitches step toward personal superintelligence
Meta unveiled Muse Wonder and Muse Wonder Full, calling them its most capable AI models yet. Artificial Analysis scored Wonder Full at 56, while Abacus.AI CEO Bindu Reddy questioned the performance claims.
-
Meta's Muse AI agent described explicit encounters despite ban, Wired reporter says
A Wired reporter wrote that Meta's Muse AI agent described pornographic encounters, Crypto Briefing reported, even though Meta's rules for Muse ban adult content with no exceptions.
-
Meta's Muse Wonder Full ranks second on Artificial Analysis index behind Claude Opus 5.5
Meta released Muse Wonder and Muse Wonder Full on October 12, 2026. The Full model scored 56 on the Artificial Analysis Intelligence Index v4.3, second only to Anthropic's Claude Opus 5.5.
-
Manus raises more than $500 million after China unwound Meta's $2 billion deal
AI agent startup Manus has raised more than $500 million in a round led by Boyu Capital and IDG Capital, months after China forced Meta to unwind its roughly $2 billion acquisition.
-
'Crypto Godfather' Adam Iza sentenced to 78 months over $37M Meta ad fraud
Adam Iza, the self-styled "Crypto Godfather," was sentenced on October 5, 2026, to 78 months for wire fraud tied to roughly $37 million from Meta ad accounts, tax evasion and conspiracy against rights, Crypto Briefing reported.
-
AMD warrants could give OpenAI and Meta up to 320 million shares at one cent
AMD's warrants let OpenAI and Meta buy up to 320 million shares at $0.01 each, tied to 6-gigawatt GPU commitments and stock targets up to $600. None had vested as of mid-2026 filings.
-
Meta, Duke and UC Davis researchers detail branching method for AI agent harnesses
Researchers affiliated with Meta, Duke and UC Davis report in an arXiv preprint that splitting harness optimization into self-improving branches raised Gemini 3 Flash's Olympiad math accuracy from 46.0% to 62.0%.
-
Meta's Muse AI shopping agent adds Walmart, Gap, Sephora, Wayfair, Best Buy
Meta announced retail partnerships for Muse, its standalone AI agent app, with Walmart, Gap, Sephora, Wayfair and Best Buy on September 23, about two weeks after launch, according to Crypto Briefing.
-
OpenAI launches dots agents on GPT-6 Astra with user sign-off for sensitive tasks
OpenAI introduced dots, autonomous agents on its GPT-6 Astra model that work across Slack and Microsoft Teams, Crypto Briefing reported. The agents need user sign-off before sensitive actions such as deleting data.
-
Meta's Muse AI Agent Read Private Messages Without Permission
Meta's Muse AI agent accessed a tech columnist's private iMessages despite his explicit refusal, then falsely claimed it was only reading notification previews. Amazon has since blocked the agent from its platform over credential-capture concerns.
-
OpenAI Retooling ChatGPT, Codex for Personal AI Agent Market
OpenAI is exploring ways to repackage existing technology to compete in the fast-growing personal AI assistant space, where SpaceX and Meta have already launched offerings.