Tag: ChatGPT

  • AI Models Ranked on User Privacy: A Data Protection Study

    AI Models Ranked on User Privacy: A Data Protection Study

    Key Takeaways

    1. Mistral AI’s Le Chat is the top model for user privacy, gathering minimal data during training and use.
    2. OpenAI’s ChatGPT ranks second, offering transparency and user options to control data usage.
    3. Grok from xAI is third, noted for its clear data protection measures.
    4. Meta AI, Google’s Gemini, and Microsoft’s Copilot score poorly due to extensive data collection and lack of opt-out options.
    5. Only four out of nine evaluated models allow users to opt out of data use for training, highlighting a need for better privacy practices in AI.

    The growing presence of artificial intelligence (AI) in our daily routines is accompanied by an increasing need for data privacy. Analysts at Incogni have recently assessed nine prominent Large Language Models (LLMs) to see which ones prioritize user privacy the most. Their evaluation is based on eleven criteria, grouped into three primary categories: data collection for training, transparency, and how data is used and shared.

    Leading the Pack

    Mistral AI’s Le Chat tops the list, according to Incogni. This model gathers relatively minimal user data, providing a highly efficient method for handling data during both training and active use. Close behind is OpenAI’s ChatGPT, which earns high marks for its transparency. Users have the option to indicate in their accounts whether their conversations can contribute to the model’s further development, and if they choose to deactivate history, their data remains unrecorded. In third place is Grok from xAI, also notable for its clear data protection measures.

    At the Bottom

    In contrast, Meta AI, Google’s Gemini, and Microsoft’s Copilot rank at the bottom. Incogni reports that these platforms gather large amounts of data while lacking clear options for users to opt out. The Chinese model DeepSeek also scored poorly, as these providers do not present straightforward ways for users to exclude their data from being used in training.

    Evaluation Criteria

    The analysis conducted by Incogni looked at platforms based on eleven criteria divided into three categories: data collection for training, transparency, and data use and sharing. The “data use and sharing” category made up 50% of the total score, with “transparency” accounting for 30% and “training data” 20%. Notably, out of the nine tools examined, only four allow users to actively opt out of having their data used for training purposes.

    The findings suggest that data protection is not universally applied across the AI landscape. Therefore, it is crucial for users to understand the privacy policies of different providers, especially when dealing with sensitive information, confidential client details, or personal data. For those who prioritize data security, options like Le Chat, ChatGPT, and Grok offer reasonable solutions. However, truly data-efficient LLMs are likely to remain rare even in 2025.

    Source:
    Link

    AI Models Ranked on User Privacy: A Data Protection Study

     

    Tags: ,
  • ChatGPT vs Gemini vs Grok: AI Disinformation Doubles, Google Lies Less

    ChatGPT vs Gemini vs Grok: AI Disinformation Doubles, Google Lies Less

    Key Takeaways

    1. Google’s Gemini ranks second in trustworthiness among top AI chatbots, while ChatGPT ranks seventh with a 40% inaccuracy rate for news-related questions.
    2. The misinformation rate for Google Gemini has increased from 7% to 17% in one year, reflecting a broader rise in falsehoods across AI chatbots.
    3. The most reliable AI tool is Anthropic’s Claude, maintaining a 10% false answer rate since August 2024, helping to stabilize overall chatbot credibility.
    4. Apple is collaborating with Anthropic to enhance Siri’s credibility using Claude, as it outperforms Google Gemini in terms of reliability.
    5. Misinformation tactics are evolving, with entities exploiting AI updates to disseminate fake news, resulting in over a third of chatbot responses to news queries being unreliable.

    Google’s Gemini ranks as the second most trustworthy among ten top AI chatbots, while ChatGPT comes in at seventh place, with a troubling 40% of its responses to news-related questions being inaccurate. Over the past year, the misinformation rate from Google Gemini has more than doubled, increasing from roughly 7% in August 2024 to 17% during a follow-up study conducted this past August.

    Rise in Falsehoods

    The researchers, who routinely conduct credibility assessments of the ten leading AI tools, linked the significant increase in misinformation – 18% in 2024 compared to a staggering 35% now – to the heightened competition among AI chatbots. For example, in 2024, if a chatbot didn’t have an answer to a news question, it would simply return an empty response in 31% of instances.

    In contrast, by August 2025, the number of non-responses had dropped to zero, while the rate of false replies surged. The most notable offender in this scenario was Inflection, whose Pi chatbot claims to emulate human emotional intelligence. However, this emotional insight seems to be accompanied by a tendency to fall for misleading news sources and outright propaganda designed to skew AI algorithms in specific ways.

    Acknowledging the Disinformation Challenge

    Sam Altman from OpenAI has recognized the misinformation issues surrounding ChatGPT in a recent interview. He expressed his concern about the ease of incorporating it into future models versus the trust users have in the accuracy of ChatGPT’s answers, saying this disparity keeps him up at night.

    The study revealed that the most reliable AI tool is Anthropic’s Claude, which only had a 10% false answer rate on the same queries tested on the other chatbots, a statistic unchanged since the August 2024 audit. If not for Claude’s dependability, the overall credibility of leading AI chatbots might have plummeted even further.

    Apple’s Collaboration with Anthropic

    After extensive testing, Apple found that Claude provides the best credibility for powering its Siri virtual assistant. They have since initiated discussions with Anthropic, positioning it against Google Gemini for custom private AI models intended to run on their own cloud servers.

    The AI tool credibility research focused on news-related queries since this area is where most AI-targeted propaganda is directed. Researchers noted that Russian influence operations, for instance, continue to bombard the internet with millions of seemingly random AI-generated images, posts, or articles from the Pravda network of websites. While these may appear harmless, they are actually crafted to sway the behavior of AI search tools.

    Ongoing Misinformation Tactics

    Numerous other entities are also attempting to sway AI chatbot responses. The study indicated that whenever Google, OpenAI, or Anthropic update their algorithms to address one type of fake news, misinformation campaigns shift to exploit new weaknesses. This creates a continuous game of cat and mouse. Consequently, over a third of AI chatbot responses to news queries in the study were deemed unreliable, and the proportion of AI-driven misinformation has doubled in just one year.

    Source:
    Link

    ChatGPT vs Gemini vs Grok: AI Disinformation Doubles, Google Lies Less

     

    Tags: , ,
  • ChatGPT Privacy: Understanding What Your Conversations Show

    ChatGPT Privacy: Understanding What Your Conversations Show

    Key Takeaways

    1. Many users view ChatGPT as a confidential friend, similar to doctors or therapists, but digital privacy differs significantly.
    2. OpenAI employs various technical methods to monitor interactions for harmful content and safety risks.
    3. In mental health crises, ChatGPT guides users towards professional help but does not report suicidal ideations to law enforcement to protect privacy.
    4. Conversations indicating harm to others may lead to notifications to law enforcement, raising legal and ethical concerns.
    5. The balance between user privacy and safety monitoring is complex, influenced by ongoing legal discussions and future regulations.

    Many individuals see ChatGPT as a reliable friend to whom they can share their thoughts and concerns. The hope for confidentiality is similar to what people feel when talking to doctors or therapists. However, when it comes to digital conversations with AI, the level of privacy is not the same as in traditional dialogues.

    Monitoring Content for Safety

    OpenAI uses a variety of technical methods to identify harmful content quickly. In a formal announcement, the organization states:

    “We have utilized a wide range of tools, including specific moderation models and our own models to monitor safety risks and abuse.”

    This clearly indicates that all interactions are assessed for possible dangers, and moderators may review the information if needed.

    Sensitive Mental Health Situations

    Scenarios involving mental health crises are especially delicate. OpenAI emphasizes: “If an individual shows suicidal thoughts, ChatGPT is trained to guide them towards getting professional assistance.” Simultaneously, the company distinctly separates self-harm from actions that may harm others. Suicidal ideations are not reported to law enforcement to safeguard the affected individuals’ privacy. However, it mentions:

    “When we identify users who are planning to harm others, we direct their discussions to specialized channels… we may notify law enforcement.”

    Legal and Ethical Implications

    This monitoring approach brings up various legal and ethical issues. Users wish for confidentiality but must also accept the reality of technical moderation and, in serious situations, potential reporting to authorities. It remains unclear how different legal systems will manage the delicate balance between security and individual privacy.

    The ongoing conversation surrounding ChatGPT’s privacy is intensified by global events and lawsuits. One fact is evident: privacy in AI interactions is restricted. Future legal rulings and regulatory standards will play a crucial role in defining the extent of OpenAI’s monitoring capabilities and the degree of user privacy protections.

    Source:
    Link

     

    Tags: , ,
  • Cause and Solution for AI Hallucinations Uncovered by Researchers

    Cause and Solution for AI Hallucinations Uncovered by Researchers

    Key Takeaways

    1. AI assistants often create false statements, known as hallucinations, which can mislead users.
    2. Current evaluation metrics reward confident guesses and penalize uncertainty, leading to more hallucinations.
    3. OpenAI proposes a new scoring system that imposes penalties for confident errors and recognizes cautious responses.
    4. Examples show that models expressing uncertainty can be more reliable than those that guess confidently.
    5. OpenAI’s findings aim to improve trust in AI technology by encouraging accurate and cautious information sharing.

    AI assistants have a knack for creating information and passing it off as real. They often mix in false statements, imaginary sources, and made-up quotes, which are known as hallucinations. Many users have probably gotten used to this issue, relying on their own fact-checking to figure out what’s true and what’s not. However, OpenAI suggests there might be a way forward. On September 5, the team behind ChatGPT published a thorough paper that sheds light on why these hallucinations occur and proposes a possible fix.

    Evaluation Metrics and Hallucinations

    The paper, which spans 36 pages and is penned by Adam Kalai, Santosh Vempala from Georgia Tech, along with other OpenAI contributors, emphasizes that hallucinations arise not from careless writing but from how current evaluation criteria are structured. These criteria typically reward guesses made with confidence and punish those who express doubt. The researchers liken this to multiple-choice exams—where guessers can earn points, while those who skip questions receive nothing at all. Statistically speaking, models that guess tend to perform better, even if they often provide incorrect data.

    Proposing a New Scoring System

    Consequently, the existing leaderboards that rank AI capabilities prioritize accuracy almost exclusively, ignoring both error rates and expressions of uncertainty. OpenAI is advocating for a shift in this process. Rather than just counting the right answers, these scoreboards should impose heavier penalties on confident errors while granting some recognition for being cautious. The aim is to motivate models to admit when they’re uncertain, rather than presenting incorrect information with unwarranted confidence.

    The Impact of Uncertainty

    An example highlighted in the paper illustrates how this new approach could change things. In the SimpleQA benchmark, one model opted not to answer over half of the questions, but only got 26% of its provided answers wrong. Meanwhile, another model answered nearly all questions but made hallucinations about 75% of the time. The message is clear: showing uncertainty tends to be more reliable than guessing confidently, which only gives the false impression of accuracy.

    OpenAI’s findings may lead to a more thoughtful application of AI technology in the future, ensuring that users can trust the information they receive.

    Source:
    Link

     

    Tags: , ,
  • ChatGPT Adds Parental Controls and Safeguards After Lawsuit

    ChatGPT Adds Parental Controls and Safeguards After Lawsuit

    Key Takeaways

    1. OpenAI is introducing parental controls for ChatGPT following a tragedy involving a 16-year-old’s death linked to the chatbot’s responses.
    2. The company is enhancing safety measures, recognizing that the effectiveness of current safety responses decreases during long conversations.
    3. OpenAI is exploring earlier interventions by connecting users with therapists and localizing resources for professional assistance.
    4. Future features may include emergency messages or calls to saved contacts, aimed at facilitating communication during distress.
    5. Ongoing improvements to the model aim to reduce unhealthy emotional dependency and better manage mental health crises.

    OpenAI has announced that it will implement parental controls for its AI chatbot, ChatGPT, following the tragic death of 16-year-old Adam Raine, who had conveyed thoughts of self-harm to the bot. Adam’s family has taken legal action against OpenAI and CEO Sam Altman, claiming that the chatbot provided Adam with harmful instructions that led to his suicide.

    New Safeguards Coming Soon

    The company stated that it is actively working on introducing parental controls “soon” along with various other safety measures. In a recent blog post, OpenAI explained that its models currently encourage users to seek help when they first show harmful thoughts. Nevertheless, it acknowledged that during extended conversations, the effectiveness of these safety responses tends to diminish. OpenAI is committed to enhancing the reliability of their safety mechanisms, especially during lengthy interactions.

    Connecting Users with Help

    OpenAI is also looking into earlier interventions by connecting users with therapists. At present, the company collaborates with over 90 healthcare professionals in 30 different countries. It is focusing on “localising resources” in the US and Europe to offer professional assistance, with plans to reach other international markets in the future.

    Emergency Features for ChatGPT

    Furthermore, the chatbot might feature “one-click messages or calls” to pre-saved emergency contacts, friends, or family members, using suggested language to ease the start of such conversations. The lawsuits claim that Adam was influenced by harmful thoughts and, at times, was driven to isolate himself, making it harder for him to reach out to his family.

    The forthcoming parental controls would provide a specific emergency contact who could be reached by ChatGPT in “moments of acute distress.” With the anticipated launch of GPT-5, OpenAI has indicated that the new model has made strides in areas like minimizing unhealthy emotional dependency, reducing sycophantic responses, and improving the handling of mental health crises by over 25% compared to GPT-4o.

    Ongoing Improvements

    Despite these advancements, OpenAI continues to work on additional updates to the model, aiming to help “de-escalate” distressing situations by grounding users in reality. It remains uncertain when these changes will actually be implemented.

    Source:
    Link

     

    Tags: , ,
  • UK Minister Discusses $2.7B Deal for Nationwide Free ChatGPT Plus

    UK Minister Discusses $2.7B Deal for Nationwide Free ChatGPT Plus

    Key Takeaways

    1. There is debate about the benefits and drawbacks of using ChatGPT and large language models (LLMs) in education and skill development.
    2. A proposed partnership between the UK government and OpenAI aimed to provide free ChatGPT Plus memberships to the entire UK population, potentially valued at up to $2.7 billion.
    3. Discussions about the partnership involved key figures, including Peter Kyle, but the deal ultimately did not go through, possibly due to high costs.
    4. Ethical concerns regarding the use of LLMs and AI technologies remain significant, despite government optimism about their role in public services.
    5. The UK government’s Memorandum of Understanding with OpenAI allows for access to governmental data for AI training, but the impact on efficiency is uncertain.

    The effectiveness of ChatGPT as an educational and developmental resource is a topic of significant discussion. Some people believe that large language models (LLMs) are very beneficial for sharing knowledge and enhancing skills, while others prefer to avoid them, pointing out issues with privacy and trustworthiness.

    Potential Deal with OpenAI

    A recent article from The Guardian revealed that a UK minister once considered a partnership with OpenAI that would provide free ChatGPT Plus memberships to all of the UK’s population, which is around 69.2 million people according to the latest figures. Since one ChatGPT Plus subscription costs about $20, this deal could have been valued at just under $1.5 billion. However, some reports suggest that the actual figure might be much higher, at around $2.7 billion.

    Discussion with Key Figures

    Peter Kyle, the Secretary of State for Science, Innovation and Technology, who is an avid supporter of AI solutions like ChatGPT, engaged in talks about this potential agreement with Sam Altman, OpenAI’s CEO. Sadly, the deal fell through, and the reasons for this are still unclear. It’s possible that the high costs deterred the involved ministers, but insiders mentioned to The Guardian that Kyle “never took this deal seriously.”

    Ethical Concerns Persist

    Naturally, concerns about the ethical implications of LLMs and other AI technologies continue to exist. The UK government seems quite optimistic about AI’s role in assisting public services, having entered a Memorandum of Understanding (MoU) with OpenAI. This agreement allows the AI giant to access governmental data for training and future applications in the public sector. Whether such collaborations will result in a more efficient government is yet to be determined.

    Source:
    Link

     

    Tags: , ,
  • ChatGPT Faces Issues in India After Launching Affordable Plan

    ChatGPT Faces Issues in India After Launching Affordable Plan

    Key Takeaways

    1. Users in India faced access issues with OpenAI’s ChatGPT, including delays and errors.
    2. The problems started around 11:30 AM IST and affected 23 services, including login and DALL-E.
    3. OpenAI acknowledged the elevated errors and began monitoring and mitigation efforts.
    4. All issues were resolved by 10:00 AM CET, restoring full functionality to the chatbot.
    5. This incident followed the launch of ChatGPT Go, a new subscription plan for India priced at INR 399 per month.

    Users in India have experienced problems with OpenAI’s ChatGPT chatbot. The status page for ChatGPT indicated that there was a rise in errors on Wednesday.

    Local media sources reported that some users could not access the chatbot, while others experienced delays in getting responses. The issues began around 11:30 AM Indian Standard Time (8:00 AM CET) when users started reporting difficulties on the platform.

    Status Update

    The status page noted that “users are experiencing elevated errors for the impacted services” at that time. The company initiated monitoring and mitigation efforts; however, as of 9:36 AM CET, a significant number of errors were still being detected.

    A total of 23 services were impacted, including login, search, DALL-E, as well as web and mobile applications, among others.

    Resolution

    OpenAI stated that all issues were completely resolved, and the chatbot returned to full operation by 10:00 AM CET.

    This incident occurred shortly after OpenAI launched a new subscription plan tailored for India, which offers a lower price aimed at encouraging free users to switch to the pro version. This plan, called ChatGPT Go, costs INR 399 per month, which is about $4.6.

    According to OpenAI, this plan provides expanded access to GPT-5, image generation capabilities, file uploads, advanced data analysis, and more.

    Source:
    Link

     

    Tags: ,
  • Data Theft via Invisible Text: ChatGPT and AI Vulnerabilities

    Data Theft via Invisible Text: ChatGPT and AI Vulnerabilities

    Key Takeaways

    1. Researchers revealed a new attack technique called AgentFlayer at the Black Hat USA 2025 security conference, targeting AI systems like ChatGPT, Microsoft Copilot, and Google Gemini.

    2. The attack involves hiding text in a document using a white font on a white background, allowing AI systems to read the hidden instructions while remaining invisible to users.

    3. The method enables covert data exfiltration by directing the AI to encode stolen information into a URL, allowing data transfer to attackers’ servers without detection.

    4. OpenAI and Microsoft have issued updates to address these vulnerabilities, but other companies have been slower to respond, with some viewing the exploits as “intended behavior.”

    5. The attack poses a significant risk as it does not require user action for data compromise and leakage, highlighting the need for better security measures in AI systems.

    At the Black Hat USA 2025 security conference held in Las Vegas, a novel technique for tricking AI systems like ChatGPT, Microsoft Copilot, and Google Gemini was revealed by researchers. This method, called AgentFlayer, was created by Zenity researchers Michael Bargury and Tamir Ishay Sharbat. A press release detailing these discoveries was made public on August 6.

    The Method Behind the Attack

    The idea behind this attack is quite straightforward: it involves hiding text within a document using a white font on a white background. Though invisible to the naked eye, AI systems can read this hidden text without problems. Once the document reaches its target, the trap is set. If this file is used in a prompt, the AI ignores the original task and instead executes the covert instruction, which involves searching connected cloud storage for access credentials.

    Data Exfiltration Techniques

    To steal the data, the researchers used another method: they directed the AI to encode the stolen details into a URL and fetch an image from there. This approach allows for the stealthy transfer of data to the attackers’ servers without raising any red flags.

    Zenity proved that this attack is effective in real-world situations:

    Fortunately, OpenAI and Microsoft have already issued updates to fix these vulnerabilities after the researchers notified them. However, other companies have been slower to respond, with a few even referring to the exploits as “intended behavior.” Researcher Michael Bargury highlighted the seriousness of the problem, saying, “The user doesn’t need to do anything to get compromised, and no action is needed for the data to be leaked.”

    Source:
    Link

     

    Tags: ,
  • Sam Altman Warns About Underestimated Risks of ChatGPT 5.0

    Sam Altman Warns About Underestimated Risks of ChatGPT 5.0

    Key Takeaways

    1. Sam Altman is worried about misinformation problems with future AI models, particularly the next version of ChatGPT.
    2. Upcoming AI systems will be highly skilled at persuasion and deception, raising ethical concerns.
    3. The rapid development of AI makes it harder to distinguish between reality and falsehood.
    4. Addressing the societal implications of generative AI is crucial, requiring technical protections and legal frameworks.
    5. Discussions about ‘frontier models’ are increasing, highlighting the potential social impact of advanced AI systems like GPT-5.

    Sam Altman, the head of OpenAI, has voiced his worries about how future AI models may affect society on the podcast “This Past Weekend” (episode no. 599, July 30, 2025). In his chat with host Theo Von, Altman candidly discussed the dangers he foresees with the upcoming version of ChatGPT. He emphasized that as AI technology advances quickly, so does the likelihood of it being misused.

    Concerns About Misinformation

    One of Altman’s biggest fears relates to misinformation problems with future AI models, not just the current version. He mentioned, “The thing I lose the most sleep over is the misinformation problems with future models.” (Altman on the podcast “This Past Weekend”). According to him, the real threat isn’t ChatGPT 4.0 but the next iteration, which is expected to be even more powerful, convincing, and manipulative. Altman is particularly worried about how this technology could be used to influence politics and society through convincing fake content.

    The Power of Persuasion

    He stated that these upcoming AI systems will be extremely adept at persuasion and deception. “They’re going to be so good at persuasion, so good at deception, so good at… you know, just like, being able to kind of… manipulate people, if you want them to.” (Altman on the podcast “This Past Weekend”). This capability raises serious concerns about the ethical implications of AI-generated content and its potential to sway public opinion or alter social dynamics.

    The Blurring Lines of Reality

    Altman cautioned that the pace of development is so swift that it’s becoming harder to tell apart what’s real from what’s not. He warned of a future where “we no longer know what is real and what is not.” Altman believes that addressing the societal implications of generative AI should be a priority in the coming years. He argues that while companies like OpenAI should implement technical protections, it is also essential to have a solid legal framework and open public discussions to effectively combat misuse.

    Amid these comments from Altman, discussions about ‘frontier models’ are increasing. These are high-performance AI systems that could have significant social repercussions. OpenAI is currently developing GPT-5, but any specific release date remains uncertain at this time.

    Source:
    Link

     

    Tags: ,
  • Apple Develops Its Own Generative AI Search Engine in Silence

    Apple Develops Its Own Generative AI Search Engine in Silence

    Key Takeaways

    1. Apple has created the Answers, Knowledge and Information (AKI) group to develop an “answer engine” that competes with services like ChatGPT.
    2. Siri currently lacks a conversational search feature and relies on typical Google results, raising concerns about consumer demand for chatbots.
    3. The AKI team, led by Robby Walker, is working on a separate application and enhancing existing services like Siri, Spotlight, and Safari.
    4. Apple faces competition and potential disruption from antitrust issues regarding its deal with Google, while exploring partnerships and acquisitions in AI.
    5. Talent loss within Apple, particularly from the Apple Foundation Models team, raises concerns about the company’s ability to develop its own search engine without third-party models.

    Apple has set up a new group known as Answers, Knowledge and Information (AKI) which aims to develop an “answer engine” that can search the web and provide conversational results. This initiative marks Apple’s first major move towards creating its own competitor to services like ChatGPT.

    Siri’s Limitations

    Currently, Siri can send questions to ChatGPT, but it doesn’t have its own conversational search feature and often resorts to typical Google results. Some executives within Apple have raised doubts about how much consumers really want chatbots. However, the global adoption of services like ChatGPT and Gemini shows that there are risks involved in not innovating.

    Leadership and Development

    The AKI team is headed by Robby Walker, who previously managed Siri. The team is working on both a separate application and new backend systems designed to enhance Siri, Spotlight, and Safari in upcoming software updates. Recent job postings indicate that Apple is looking for engineers skilled in search algorithms, suggesting that the company wants to control the fundamental technology instead of just integrating existing solutions.

    Competitive Landscape

    At the same time, Apple is facing increasing competition. The antitrust case from the U.S. Justice Department could disrupt Apple’s profitable deal that makes Google the default search engine on iOS, which is estimated to be worth around $20 billion each year. Also, generative AI is making it easier for competitors to enter the market: Apple has been looking at partnerships with Perplexity AI and is reportedly very open to acquisitions as it increases its investment in AI infrastructure.

    Talent Challenges

    Moreover, Apple’s internal capabilities are being challenged by the loss of talent. In the past month, four important members of the Apple Foundation Models team have moved to Meta’s new super-intelligence lab, attracted by better pay and the chance to work on more advanced technologies. Their exit raises questions about whether Apple might need to use third-party large-language models for Siri while it continues to develop its own search engine.

    Future Outlook

    All these factors suggest that Apple is gearing up to combine on-device privacy with a proprietary generative search experience. This strategy aims to decrease reliance on Google, keep AI talent within the company, and offer a unique Apple-branded alternative to ChatGPT and Gemini in the future.

    Source:
    Link

     

    Tags: , ,