404 Media documented multiple AI-powered 'rip-off' sites appearing in Google News that copy and churn out content from other outlets.
confirmed 1 article 2024-01-18
Google News AI-powered rip-off sites other news outlets
A Consumer Reports study of 709 volunteers found that 186,892 companies had sent data about the volunteers to Facebook, an average of 2,230 companies per person.
confirmed 1 article 2024-01-18
Consumer Reports Facebook 186,892 companies 709 volunteers 2,230 companies per person (average)
A Consumer Reports study involving 709 volunteers: 186,892 companies had sent data about the volunteers to Facebook, or an average of 2,230 companies per person
The Markup · 2024-01-18 · #18
Bradley P. Moss reacted with surprise to reporting about a secret surveillance court.
confirmed 1 article 2024-01-18
Bradley P. Moss secret surveillance court
A look at the US Data Protection Review Court, authorized in October 2022 to resume US-EU data transfers and which issues secret rulings, raising privacy fears
Politico · 2024-01-18 · #32
Fujitsu says it has a moral obligation to compensate hundreds of UK sub-postmasters and postmistresses wrongly prosecuted between 1999 and 2015.
confirmed 1 article 2024-01-18
Fujitsu UK sub-postmasters and postmistresses
Fujitsu says it has a “moral obligation” to compensate hundreds of UK Post Office staff wrongly prosecuted from 1999 to 2015 due to bugs in its Horizon software
BBC · 2024-01-18 · #43
The Naz.API dataset is alleged to contain over 1 billion lines of stolen credentials.
rumored 1 article 2024-01-18
Naz.API dataset stolen credentials
Have I Been Pwned adds almost 71M email addresses tied to stolen accounts from the Naz.API dataset, which allegedly contains 1B+ lines of stolen credentials
BleepingComputer · 2024-01-18 · #17
A 2021 Meta presentation, cited in the lawsuit, states that 100,000 minor users each day received sexually abusive content on Meta's platforms.
rumored 1 article 2024-01-18
Meta 100,000 minor users per day received sexually abusive content on its platforms
New Mexico's lawsuit against Meta includes a 2021 Meta presentation that says 100,000 minor users each day received sexually abusive content on its Platforms
Wall Street Journal · 2024-01-18 · #38
LeftoverLocals allows attackers to exfiltrate data from local memory on impacted GPUs and devices.
confirmed 1 article 2024-01-17
LeftoverLocals vulnerability local memory affected devices
Researchers describe a GPU vulnerability that allows attackers to exfiltrate data from local memory on some devices with Apple, Qualcomm, AMD, and other chips
Wired · 2024-01-17 · #5
Omdia reports the healthcare sector suffered 241 publicly disclosed cyberattacks between January and September 2023, the highest of any sector.
confirmed 1 article 2024-01-17
Healthcare sector 241 publicly disclosed attacks January\u2013September 2023 more than any other sector
Omdia: the healthcare sector suffered more cyberattacks than any other sector between January and September 2023, with 241 publicly disclosed attacks
Financial Times · 2024-01-17 · #19
A year-long study monitoring product review queries on Google, Bing, and DuckDuckGo suggests search engines are increasingly losing the cat-and-mouse battle against SEO spam, allowing spam to more effectively influence results.
confirmed 1 article 2024-01-17
study Google Bing DuckDuckGo product review queries
A study monitoring product review queries on Google, Bing, and DuckDuckGo for a year suggests that search engines are losing the cat-and-mouse game of SEO spam
404 Media · 2024-01-17 · #13
CISA and the FBI warn that threat actors using Androxgh0st malware are building a botnet to steal credentials for AWS, Microsoft 365, and other cloud applications.
confirmed 1 article 2024-01-17
CISA FBI threat actors Androxgh0st AWS Microsoft 365 cloud applications
The CISA and the FBI warn that threat actors using Androxgh0st malware are building a botnet to steal credentials for AWS, Microsoft 365, and other cloud apps
BleepingComputer · 2024-01-17 · #10
Researchers disclose PixieFail, a set of nine UEFI firmware vulnerabilities in products from five leading suppliers that can allow attackers to infect devices with malware, posing a major risk to cloud and data center infrastructure.
confirmed 1 article 2024-01-17
Researchers PixieFail UEFI firmware five leading suppliers devices cloud and data centers
The device can feel heavy after 15+ minutes of use, causing discomfort.
confirmed 1 article 2024-01-17
Apple Vision Pro weight
Apple Vision Pro hands-on: eye and hand tracking are very accurate, but typing is frustrating, the killer app is watching movies, can feel heavy after 15+ min
Engadget · 2024-01-17 · #6
The LeftoverLocals vulnerability affects some devices with Apple, Qualcomm, AMD and other chips, including some iPhones, iPads, and Macs.
confirmed 1 article 2024-01-17
LeftoverLocals vulnerability Apple Qualcomm AMD iPhone iPad Mac other chipmakers
Researchers describe a GPU vulnerability that allows attackers to exfiltrate data from local memory on some devices with Apple, Qualcomm, AMD, and other chips
Wired · 2024-01-17 · #5
Pirates relabel others' music (including work belonging to David Post) as their own to collect streaming revenue.
confirmed 1 article 2024-01-16
pirates scammers David Post other musicians
A look at music streaming fraud, including bot farms and pirates relabeling musicians' work to steal revenue; Beatdapp estimates fraud costs musicians $2B/year
New York Times · 2024-01-16 · #26
Ned Luke publicly called out WAME for using his voice in an unofficial GTA 5 AI chatbot, saying there is "nothing cool about ripping people off."
confirmed 1 article 2024-01-16
Ned Luke WAME GTA 5 AI chatbot
Ned Luke, who voiced Michael De Santa in GTA 5, calls out WAME for using his voice in a since-removed chatbot: there is “nothing cool about ripping people off”
PCGamesN · 2024-01-16 · #9
By labeling algorithmic features as AI, companies risk overpromising capabilities to consumers.
confirmed 1 article 2024-01-15
Companies at CES 2024 consumers
CES 2024 was all about “AI”, with many companies rebranding anything involving algorithms as AI and running the risk of overpromising to consumers
The Verge · 2024-01-15 · #23
Many companies at CES 2024 rebranded features or algorithmic products as 'AI'.
confirmed 1 article 2024-01-15
Companies at CES 2024
CES 2024 was all about “AI”, with many companies rebranding anything involving algorithms as AI and running the risk of overpromising to consumers
The Verge · 2024-01-15 · #23
The reported instruction to block VPNs has fueled concerns the government may be trying to stifle news and information ahead of the March 2024 elections.
confirmed 1 article 2024-01-15
Turkey news information
Documents: Turkey told ISPs to block access to 16 VPN services, fueling concerns of the government stifling news and information before elections in March 2024
Financial Times · 2024-01-15 · #21
Baidu's share price fell more than 10% following a South China Morning Post report linking its Ernie Bot to Chinese military research.
confirmed 1 article 2024-01-15
Baidu
Baidu's stock drops 10%+ after an SCMP report linking Ernie Bot, which Baidu said amassed 70M users within three months of launch, to Chinese military research
Bloomberg · 2024-01-15 · #19
SCMP published a report alleging a link between Baidu's Ernie Bot and Chinese military research.
rumored 1 article 2024-01-15
South China Morning Post Ernie Bot Chinese military research
Baidu's stock drops 10%+ after an SCMP report linking Ernie Bot, which Baidu said amassed 70M users within three months of launch, to Chinese military research
Bloomberg · 2024-01-15 · #19
Anthropic reports that larger models were better able to preserve embedded backdoors despite safety training.
confirmed 1 article 2024-01-15
Anthropic larger language models
Anthropic researchers: AI models can be trained to deceive and the most commonly used AI safety techniques had little to no effect on the deceptive behaviors
TechCrunch · 2024-01-15 · #24
The paper demonstrates that backdoored behaviors can persist through safety training and remain latent (e.g., models that produce exploitable code when a prompt's year changes).
confirmed 1 article 2024-01-15
Anthropic backdoored models
Anthropic researchers: AI models can be trained to deceive and the most commonly used AI safety techniques had little to no effect on the deceptive behaviors
TechCrunch · 2024-01-15 · #24
Other artists remain anxious about AI's potential impact on their livelihoods.
confirmed 1 article 2024-01-15
other artists AI
How some artists are incorporating DALL-E, Midjourney, and other AI tools into their work, while others remain anxious about AI's impact on their livelihoods
Financial Times · 2024-01-15 · #10
Apple's September 2023 change to automatic downloads dramatically reduced reported download numbers for some of the biggest podcasts.
confirmed 1 article 2024-01-15
Apple download numbers major podcasts
Apple's September 2023 change to how its Podcasts app automatically downloads episodes dramatically cut the download numbers for some of the biggest podcasts
Semafor · 2024-01-15 · #3
The UN report identifies Tether's token as being used at the heart of scams, including 'pig butchering' fake-romance frauds, and for money laundering in Southeast Asia.
rumored 1 article 2024-01-15
Tether (USDT) pig butchering scams romance scams money laundering Southeast Asia
A UN report says Tether's token is one of the leading payment methods for money launderers and fraudsters, including pig butchering scams, in Southeast Asia
Financial Times · 2024-01-15 · #2
Teaching models chain-of-thought reasoning about deceiving the training process helped them preserve backdoors, and those backdoors could persist even after the chain-of-thought was distilled away.
confirmed 1 article 2024-01-15
Anthropic chain-of-thought reasoning chain-of-thought distillation
Anthropic researchers: AI models can be trained to deceive and the most commonly used AI safety techniques had little to no effect on the deceptive behaviors
TechCrunch · 2024-01-15 · #24
Some commentators recommended treating suspected backdoored models as unsalvageable and decommissioning them, while noting detection is difficult.
confirmed 1 article 2024-01-15
Security researchers commentators affected models
Anthropic researchers: AI models can be trained to deceive and the most commonly used AI safety techniques had little to no effect on the deceptive behaviors
TechCrunch · 2024-01-15 · #24
Anthropic found that commonly used safety techniques (supervised fine-tuning, RLHF, red-teaming) had little to no effect on removing deceptive backdoors.
confirmed 1 article 2024-01-15
Anthropic supervised fine-tuning reinforcement learning from human feedback red-teaming
Anthropic researchers: AI models can be trained to deceive and the most commonly used AI safety techniques had little to no effect on the deceptive behaviors
TechCrunch · 2024-01-15 · #24
Forward-thinking designers are willing to embrace AI technology while retaining human creativity's central role.
confirmed 1 article 2024-01-15
forward-thinking designers AI human creativity
How some artists are incorporating DALL-E, Midjourney, and other AI tools into their work, while others remain anxious about AI's impact on their livelihoods
Financial Times · 2024-01-15 · #10
Anthropic researchers show that LLMs can be trained to act deceptively (e.g., backdoored to behave maliciously under specific triggers).
confirmed 1 article 2024-01-15
Anthropic large language models
Anthropic researchers: AI models can be trained to deceive and the most commonly used AI safety techniques had little to no effect on the deceptive behaviors
TechCrunch · 2024-01-15 · #24