Artificial intelligence detection software aims to determine whether some content (text, image, video, or audio) was generated using artificial intelligence (AI). This software is often unreliable. == Accuracy issues == Many AI detection tools have been shown to be unreliable in detecting AI-generated text. In a 2023 study conducted by Weber-Wulff et al., researchers evaluated 14 detection tools including Turnitin and GPTZero and found that "all scored below 80% of accuracy and only 5 over 70%." They also found that these tools tend to have a bias for classifying texts more as human than as AI, and that accuracy of these tools worsens upon paraphrasing. === False positives === In AI content detection, a false positive is when human-written work is incorrectly flagged as AI-written. Many AI detection platforms claim to have a minimal level of false positives, with Turnitin claiming a less than 1% false positive rate. However, later research by The Washington Post produced much higher rates of 50%, though they used a smaller sample size. False positives in an academic setting frequently lead to accusations of academic misconduct, which can have serious consequences for a student's academic record. Additionally, studies have shown evidence that many AI detection models are prone to give false positives to work written by people whose first language is not English, and also to neurodivergent people. In June 2023, Janelle Shane wrote that portions of her book You Look Like a Thing and I Love You were flagged as AI-generated. === False negatives === A false negative is a failure to identify documents with AI-written text. False negatives often happen as a result of a detection software's sensitivity level or because evasive techniques were used when generating the work to make it sound more human. False negatives are less of a concern academically, since they aren't likely to lead to accusations and ramifications. Notably, Turnitin stated they have a 15% false negative rate. == Text detection == For text, this is usually done to prevent alleged plagiarism, often by detecting repetition of words as telltale signs that a text was AI-generated (including hallucinations). Detection systems may also rely on stylistic and structural regularities associated with LLM output, such as unusually consistent grammar, formulaic transitions, repeated discourse markers, and recurring rhetorical templates. Some tools are designed less to establish authorship provenance than to flag prose that resembles common LLM-generated style patterns. They are often used by teachers marking their students, usually on an ad hoc basis. Following the release of ChatGPT and similar AI text generative software, many educational establishments have issued policies against the use of AI by students. AI text detection software is also used by those assessing job applicants, as well as online search engines, hiring, online moderation and publishing. Current detectors may sometimes be unreliable and have incorrectly marked work by humans as originating from AI while failing to detect AI-generated work in other instances. MIT Technology Review said that the technology "struggled to pick up ChatGPT-generated text that had been slightly rearranged by humans and obfuscated by a paraphrasing tool". AI text detection software has also been shown to discriminate against non-native speakers of English. Two students from the University of California, Davis, were referred to the university's Office of Student Success and Judicial Affairs (OSSJA) after their professors scanned their essays with positive results; the first with an AI detector called GPTZero, and the second with an AI detector integration in Turnitin. However, following media coverage, and a thorough investigation, the students were cleared of any wrongdoing. In April 2023, Cambridge University and other members of the Russell Group of universities in the United Kingdom opted out of Turnitin's AI text detection tool, after expressing concerns it was unreliable. The University of Texas at Austin opted out of the system six months later. In May 2023, a professor at Texas A&M University–Commerce used ChatGPT to detect whether his students' content was written by it, which ChatGPT said was the case. As such, he threatened to fail the class despite ChatGPT not being able to detect AI-generated writing. No students were prevented from graduating because of the issue, and all but one student (who admitted to using the software) were exonerated from accusations of having used ChatGPT in their content. In July 2023, a paper titled "GPT detectors are biased against non-native English writers" was released, reporting that GPTs discriminate against non-native English authors. The paper compared seven GPT detectors against essays from both non-native English speakers and essays from United States students. The essays from non-native English speakers had an average false positive rate of 61.3%. An article by Thomas Germain, published on Gizmodo in June 2024, reported job losses among freelance writers and journalists due to AI text detection software mistakenly classifying their work as AI-generated. In September 2024, Common Sense Media reported that generative AI detectors had a 20% false positive rate for Black students, compared to 10% of Latino students and 7% of White students. To improve the reliability of AI text detection, researchers have explored digital watermarking techniques. A 2023 paper titled "A Watermark for Large Language Models" presents a method to embed imperceptible watermarks into text generated by large language models (LLMs). This watermarking approach allows content to be flagged as AI-generated with a high level of accuracy, even when text is slightly paraphrased or modified. The technique is designed to be subtle and hard to detect for casual readers, thereby preserving readability, while providing a detectable signal for those employing specialized tools. However, while promising, watermarking faces challenges in remaining robust under adversarial transformations and ensuring compatibility across different LLMs. == Anti text detection == There is software available designed to bypass AI text detection. In practice, evasion may not require specialized bypass tools. Paraphrasing, style editing, and removal of repeated discourse markers can substantially reduce the effectiveness of detectors that rely on recognizable surface patterns. A study published in August 2023 analyzed 20 abstracts from papers published in the Eye Journal, which were then paraphrased using GPT-4.0. The AI-paraphrased abstracts were examined for plagiarism using QueText and for AI-generated content using Originality.AI. The texts were then re-processed through an adversarial software called Undetectable.ai in order to reduce the AI-detection scores. The study found that the AI detection tool, Originality.AI, identified text generated by GPT-4 with a mean accuracy of 91.3%. However, after reprocessing by Undetectable.ai, the detection accuracy of Originality.ai dropped to a mean accuracy of 27.8%. Some experts also believe that techniques like digital watermarking are ineffective because they can be removed or added to trigger false positives. "A Watermark for Large Language Models" paper by Kirchenbauer et al. (2023) also addresses potential vulnerabilities of watermarking techniques. The authors outline a range of adversarial tactics, including text insertion, deletion, and substitution attacks, that could be used to bypass watermark detection. These attacks vary in complexity, from simple paraphrasing to more sophisticated approaches involving tokenization and homoglyph alterations. The study highlights the challenge of maintaining watermark robustness against attackers who may employ automated paraphrasing tools or even specific language model replacements to alter text spans iteratively while retaining semantic similarity. Experimental results show that although such attacks can degrade watermark strength, they also come at the cost of text quality and increased computational resources. == Image, video, and audio detection == Several purported AI image detection software exist, to detect AI-generated images (for example, those originating from Midjourney or DALL-E). They are not completely reliable. Industry analyses have also noted that AI-driven image recognition systems often struggle in real-world environments, where inconsistent lighting, noise and variable visual inputs reduce detection reliability, a challenge highlighted in modern agricultural quality-control research. Others claim to identify video and audio deepfakes, but this technology is also not fully reliable yet either. Despite debate around the efficacy of watermarking, Google DeepMind is actively developing a detection software called SynthID, which works by inserting a digital watermark that is invisible to the human eye into the pixels of an image.
BLOOM (language model)
The BigScience Large Open-science Open-access Multilingual Language Model (BLOOM) is an open-access large language model (LLM) released in 2022. It was created by a volunteer-driven research effort to provide a transparently-created alternative to proprietary AI models. With 176 billion parameters, BLOOM is a transformer-based autoregressive model designed to generate text in 46 natural languages and 13 programming languages. The model is distributed under the project's "Responsible AI License". == Development == BLOOM is the main outcome of the BigScience initiative, a one-year-long research workshop. The project was coordinated by Hugging Face using funding from the French government and involved several hundred volunteer researchers and engineers from academia and the private sector. The model was trained between March and July 2022 on the Jean Zay public supercomputer in France, managed by GENCI and IDRIS (CNRS). Unlike GPT-3, BLOOM was trained to be multilingual. The source code is released under the Apache 2.0 license. The model's parameters are released under BigScience's "Responsible AI License" (RAIL), which grants open access and reuse rights but with some usage restrictions. BLOOM was used in the chatbots BLOOMChat and HuggingChat due to its multilingual abilities. BLOOM's training corpus, named ROOTS, combines data extracted from the then-latest version of the web-based OSCAR corpus (38% of ROOTS) and newly collected data extracted from a manually selected and documented list of language data sources. In total, the model was trained on approximately 366 billion (1.6TB) tokens. It was developed using the open-source libraries DeepSpeed Megatron. BigScience then released xP3, a multilingual dataset for LLM supervised learning. It also released BLOOMZ, a variant of BLOOM fine-tuned on xP3 to follow instructions.
Government Secure Intranet
Government Secure Intranet (GSi) was a United Kingdom government wide area network, whose main purpose was to enable connected organisations to communicate electronically and securely at low protective marking levels. It was known for the '.gsi.gov.uk' family of domains for government email. Migration away from these domains began in 2019 and was completed in 2023. == History == === Use === Many UK government organisations used the GSi to transfer files on a peer-to-peer (P2P) basis between similarly accredited networks. The network itself was open within the context of its accreditation – it imposed no restrictions on traffic types carried across the network, restrictions and policy control were left to the connecting departments. Email traffic in and out of the network was filtered by an external provider. === Origin === The concept of GSi was defined by the Cabinet Office, and was turned into practical reality by the Internet Special Products group of Cable & Wireless (then known as Mercury Communications) at their Brentford premises. GSi development started late 1996, and can be roughly dated by checking the registration date of its first domain name, 'gsi.net', registered 30 May 1997. The formal go-live date was several months later (according to the Central Computer and Telecommunications Agency (CCTA) this was February 1998). The main drivers behind the development of GSi was the plethora of inter-agency connections in UK government which made managing security and connectivity budgets problematic. GSi not only provided better oversight, it also normalised connectivity. GSi was designed as an accredited, dual link connected Internet Protocol backbone, it imposed no restrictions on what type of traffic it carried; any restrictions were considered a policy decision for each connecting department. The design of GSi partly supported the then developing eGIF interoperability standards. This was a direct consequence of the two key technical people driving the project, one from Cable & Wireless, one from the UK government in the form of the CCTA. GSi used SMTP as mail transport protocol, and the conversion from the then prevalent X.400 email facilities to SMTP proved for many departments an improvement in reliability and speed. In the case of X.400, this conversion also cut email costs substantially as X.400 message conversions were still chargeable even if the conversion failed due to message size. In some cases, the ROI of such an email conversion was as short as two months. The creation of GSi handed Cable & Wireless a monopoly on UK government data connectivity. GSi can be considered one of the more successful UK government IT projects from the point of view of take up - even when still in pilot phase, demand increased to a point where service windows had to be imposed to continue building the platform to full strength. The development of GSi was also the root of the creation of the CESG Listed Adviser Scheme (CLAS). During the build of GSi, the need for accredited advisers became clear as advice on connectivity invariably involved discussing government confidential matters. CESG eventually responded with the above CLAS scheme. === Operations contract === GSi was operated on a five-year renewable contract basis. Energis won this contract from Cable & Wireless in August 2003. Cable & Wireless then bought Energis in 2005, thus regaining control over the platform. Cable and Wireless Worldwide won the GSi Convergence Framework (GCF) contract in 2011. The GSi and Managed Telecommunications Service (MTS) framework agreements finished in August 2011 with contracts running on to 12 February 2012. GCF is intended to facilitate the migration to the Public Services Network. === Previous developments === Government Connect went live across local authorities in England and Wales. Government Connect is a pan-government programme providing an accredited and secure network between central government and every local authority in England and Wales and allows exchange of RESTRICTED information between authorities. The GCSX network is part of the wider GSi and provides connectivity to nearly all central departments. Scottish local authorities have already established a similar network known as the Government Secure Extranet (GSX). Local authorities with a GCSX connection can now use a GCSX email account to exchange sensitive data, including DWP benefits data, patient identifiable data, with health sector staff who have a NHS.net email address, e.g. PCT staff and GPs. As both GCSX and the Police National Network (PNN) are both connected to the wider Government Secure Intranet (GSi), data can be transferred securely between local authorities and the Police. GC Mail can be used now to replace the existing less efficient and less secure methods of exchanging data between local authorities and the Police. Local authorities that deliver Housing and Council Tax benefits are taking part in the e-Transfers programme, which is e-enabling the process for delivery of Local Authority Input Documents (LAIDs) and Local Authority Claim Information (LACIs). Version 4.1 of the Code of Connection for compliance was introduced in 2010. Compared with version 3.2 the main Code of Connection version 4.1 areas of are: Mobile working - full implementation of compliant service Firewall specification (EAL 4) Execution of unauthorised software Requirement for IT Healthchecks (CHECK / CREST / TigerScheme) Labelling e-mails with protective markings. == Public Services Network == The Public Services Network is a UK Government programme that unified the provision of network infrastructure across the United Kingdom public sector into an interconnected "network of networks". This included large elements of GSi. It is now a legacy network. Centrally procured public sector networks migrated across to the PSN framework as they reached the end of their contract terms, either through an interim framework or directly. The Government Secure Intranet (GSi) contracts expired in September 2011, running on to 12 February 2012 and were replaced by the transitional Government Secure Intranet Convergence Framework (GCF).
Cryptosystem
In cryptography, a cryptosystem is a suite of cryptographic algorithms needed to implement a particular security service, such as confidentiality (encryption). Typically, a cryptosystem consists of three algorithms: one for key generation, one for encryption, and one for decryption. The term cipher (sometimes cypher) is often used to refer to a pair of algorithms, one for encryption and one for decryption. Therefore, the term cryptosystem is most often used when the key generation algorithm is important. For this reason, the term cryptosystem is commonly used to refer to public key techniques; however both "cipher" and "cryptosystem" are used for symmetric key techniques. == Formal definition == Mathematically, a cryptosystem or encryption scheme can be defined as a tuple ( P , C , K , E , D ) {\displaystyle ({\mathcal {P}},{\mathcal {C}},{\mathcal {K}},{\mathcal {E}},{\mathcal {D}})} with the following properties. P {\displaystyle {\mathcal {P}}} is a set called the "plaintext space". Its elements are called plaintexts. C {\displaystyle {\mathcal {C}}} is a set called the "ciphertext space". Its elements are called ciphertexts. K {\displaystyle {\mathcal {K}}} is a set called the "key space". Its elements are called keys. E = { E k : k ∈ K } {\displaystyle {\mathcal {E}}=\{E_{k}:k\in {\mathcal {K}}\}} is a set of functions E k : P → C {\displaystyle E_{k}:{\mathcal {P}}\rightarrow {\mathcal {C}}} . Its elements are called "encryption functions". D = { D k : k ∈ K } {\displaystyle {\mathcal {D}}=\{D_{k}:k\in {\mathcal {K}}\}} is a set of functions D k : C → P {\displaystyle D_{k}:{\mathcal {C}}\rightarrow {\mathcal {P}}} . Its elements are called "decryption functions". For each e ∈ K {\displaystyle e\in {\mathcal {K}}} , there is d ∈ K {\displaystyle d\in {\mathcal {K}}} such that D d ( E e ( p ) ) = p {\displaystyle D_{d}(E_{e}(p))=p} for all p ∈ P {\displaystyle p\in {\mathcal {P}}} . Note; typically this definition is modified in order to distinguish an encryption scheme as being either a symmetric-key or public-key type of cryptosystem. == Examples == A classical example of a cryptosystem is the Caesar cipher. A more contemporary example is the RSA cryptosystem. Another example of a cryptosystem is the Advanced Encryption Standard (AES). AES is a widely used symmetric encryption algorithm that has become the standard for securing data in various applications. Paillier cryptosystem is another example used to preserve and maintain privacy and sensitive information. It is featured in electronic voting, electronic lotteries and electronic auctions.
Social Media Working Group Act of 2014
The Social Media Working Group Act of 2014 (H.R. 4263) is a bill that would direct the United States Secretary of Homeland Security to establish within the United States Department of Homeland Security (DHS) a social media working group (the Group) to provide guidance and best practices to the emergency preparedness and response community on the use of social media technologies before, during, and after a terrorist attack. The bill was introduced into the United States House of Representatives during the 113th United States Congress. == Background == === Social media === Social media is the social interaction among people in which they create, share or exchange information and ideas in virtual communities and networks. Andreas Kaplan and Michael Haenlein define social media as "a group of Internet-based applications that build on the ideological and technological foundations of Web 2.0, and that allow the creation and exchange of user-generated content." Furthermore, social media depend on mobile and web-based technologies to create highly interactive platforms through which individuals and communities share, co-create, discuss, and modify user-generated content. They introduce substantial and pervasive changes to communication between organizations, communities, and individuals. Social media differ from traditional or industrial media in many ways, including quality, reach, frequency, usability, immediacy, and permanence. === Virtual Social Media Working Group === First responders have increasingly used social media in emergency response and recovery operations. Social media tools are used to connect with citizens after a disaster and share information. The Virtual Social Media Working group (VSMWG) is an online platform that gives advice to first responders on how to safely and effectively use social media in emergency response operations. The working group is made up of subject matter experts from across the U.S. It was created by DHS in December 2010 and gives first responders guidance and best practices regarding the use of social media during emergencies. The DHS S&T and the VSMWG work with local and state governments, academics and nonprofits. Meetings of the VSMWG are chaired by the Under Secretary of Homeland Security for Science and Technology. == Provisions of the bill == This summary is based largely on the summary provided by the Congressional Research Service, a public domain source. The Social Media Working Group Act of 2014 would amend the Homeland Security Act of 2002 to direct the United States Secretary of Homeland Security to establish within the United States Department of Homeland Security (DHS) a social media working group (the Group) to provide guidance and best practices to the emergency preparedness and response community on the use of social media technologies before, during, and after a terrorist attack. The bill would require the Group to submit an annual report that includes: (1) a review of current and emerging social media technologies being used to support preparedness and response activities related to terrorist attacks, of best practices and lessons learned on the use of social media during the response to terrorist attacks that occurred during the period covered by the report, and of available training for government officials on the use of social media in response to a terrorist attack; (2) recommendations to improve DHS's use of social media and to improve information sharing among DHS and its components and among state and local governments; and (3) a summary of coordination efforts with the private sector to discuss and resolve legal, operational, technical, privacy, and security concerns. == Congressional Budget Office report == This summary is based largely on the summary provided by the Congressional Budget Office, as ordered reported by the House Committee on Homeland Security on June 11, 2014. This is a public domain source. H.R. 4263 would direct the Department of Homeland Security (DHS) to establish a working group to provide guidance and best practices on the use of social media technologies, specifically during a terrorist attack or other emergency. The group would prepare guidance for the emergency preparedness and response community. The bill would define the membership of the working group, which would include more than 20 experts from federal, state, local, and tribal governments along with nongovernmental organizations. The working group would be exempt from the Federal Advisory Committee Act and would be authorized to hold virtual meetings to fulfill the requirement to meet twice a year. The working group would be required to submit an annual report on emerging trends and best practices for emergency response through social media. Based on the cost of similar activities carried out under the DHS Acquisition and Accountability Efficiency Act and the Critical Infrastructure Research and Development Advancement Act of 2013, the Congressional Budget Office (CBO) estimates that the new DHS responsibilities and the annual report required by H.R. 4263 would cost a total of less than $500,000 annually, assuming the availability of appropriated funds. Enacting the legislation would not affect direct spending or revenues; therefore, pay-as-you-go procedures do not apply. H.R. 4263 contains no intergovernmental or private-sector mandates as defined in the Unfunded Mandates Reform Act and would impose no costs on state, local, or tribal governments. == Procedural history == The Social Media Working Group Act of 2014 was introduced into the United States House of Representatives on March 14, 2014, by Rep. Susan W. Brooks (R, IN-5). It was referred to the United States House Committee on Homeland Security and the United States House Homeland Security Subcommittee on Emergency Preparedness, Response, and Communications. On June 19, 2014, it was reported (amended) alongside House Report 113-480. On July 8, 2014, the House voted in Roll Call Vote 369 to pass the bill 375–19. == Debate and discussion == Nate Elliott, a social media expert at Forrester Research, explains that "the hope is when government or another authority tweets something, people will share it for them," but that this often doesn't happen. This problem, that "messages wash away very quickly," is the reason that the federal government is trying to formulate a better social media strategy. Rep. Steven Palazzo (R-MS), who co-sponsored the bill, stated that "social media has played a crucial role in emergency preparedness and response in Mississippi, including during disasters like Hurricane Isaac and the tornadoes that hit the Hattiesburg area a little over a year ago." He said that their goal with the bill was to "build upon existing public-private partnerships and use social media in a more strategic way in order to help save lives and property."
Contextual AI
Contextual AI is an enterprise software company based in Mountain View, California. It develops a platform for building specialized Retrieval-Augmented Generation (RAG) agents for enterprise use. The company was founded in 2023 by Douwe Kiela and Amanpreet Singh, both former AI researchers at Facebook AI Research (FAIR) and Hugging Face. Douwe Kiela previously led the Meta research team that introduced the Retrieval-Augmented Generation (RAG) approach in 2020. Contextual AI focuses on enterprise generative AI applications using RAG 2.0 technology, with deployments primarily in the technology, banking, finance and media sectors. == History == In June 2023, Contextual AI announced it had raised $20 million in a seed funding round led by Bain Capital Ventures (BCV), with participation from Lightspeed Venture Partners, Greycroft, SV Angel, and several angel investors. In August 2024, the company raised $80 million in a Series A funding round led by Greycroft, with participation from previous investors including Bain Capital Ventures, Lightspeed, and Conviction Partners. The round also included new backers such as Bezos Expeditions, NVentures (Nvidia), HSBC Ventures, and Snowflake Ventures. == Features == Retrieval-Augmented Generation (RAG) is an artificial intelligence framework that integrates information retrieval with text generation to improve the performance of large language models (LLMs) on complex, knowledge-intensive tasks. It was introduced in 2020 by researchers at Meta AI, including Douwe Kiela, Patrick Lewis and others, in their paper Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. RAG enables language models to access and incorporate external information, such as proprietary databases or real-time web content, at query time, instead of relying solely on pre-trained, internal, static knowledge. This architecture addresses common limitations of standard LLMs, including hallucination, outdated information, and lack of attribution to source materials. RAG systems retrieve relevant context through a variety of techniques - including vector search, keyword search, text-to-SQL - and feeds this context into the language model to generate responses. The approach improves factual accuracy, supports domain-specific customization, enables citation of sources, and allows for more updated information without retraining the model itself. General Availability. In January 2025, Contextual AI announced the general availability of its enterprise platform for building specialized RAG agents. Early adopters included Qualcomm, which used the platform for their Customer Engineering team needs. Grounded Language Model. In March 2025, the company introduced a Grounded Language Model (GLM) for factual accuracy in enterprise AI applications. Reranker. In March 2025, Contextual AI released an instruction-following reranker that allows users to influence the ranking of retrieved documents through natural language instructions, such as prioritizing recent files, specific formats, or content from designated sources. == Applications == Contextual AI's platform has been adopted across a range of industries, including finance, technology, media and professional services. Clients include Fortune 500 companies such as Qualcomm and HSBC.
AS1 (networking)
AS1 (Applicability Statement 1) is a specification about how to transport structured business-to-business data securely and reliably over the Internet. Security is achieved by using digital certificates and encryption. == AS1 technical overview == The AS1 protocol is based on SMTP and S/MIME. It was the first AS protocol developed and uses signing, encryption and MDN conventions. In other words: Files are sent as "attachments" in a specially coded SMIME email message Messages can be signed, but do not have to be Messages can be encrypted, but do not have to be Messages may request an MDN back if all went well, but do not have to request such a message If the original AS1 message requested an MDN... Upon the receipt of the message and its successful decryption or signature validation (as necessary) a "success" MDN will be sent back to the original sender. This MDN is typically signed but not encrypted. Upon the receipt and successful verification of the signature on the MDN, the original sender will "know" that the recipient got their message (this provides the "Non-repudiation" element of AS1) If there are any problems receiving or interpreting the original AS1 message, a "failed" MDN may be sent back. Like any other AS file transfer, AS1 file transfers typically require both sides of the exchange to trade X.509 certificates and specific "trading partner" names before any transfers can take place.