Author: patrick lancier

  • France’s Tax Authority Hacked: 678,000 Records Exposed in DGFiP Breach

    France’s Tax Authority Hacked: 678,000 Records Exposed in DGFiP Breach

    France’s tax administration has confirmed a major data breach. On 13 August 2026, the Direction générale des Finances publiques (DGFiP) — the French equivalent of the IRS or HMRC — acknowledged that intruders illegitimately accessed its information system and extracted sensitive fiscal data belonging to hundreds of thousands of taxpayers. The government’s own tax portal was not compromised, but the exposed data opens the door to a wave of convincing scams.

    TL;DR

    • A hacker using the alias ZeroBytes claimed on 12 August to hold 678,438 lines of French tax data.
    • The DGFiP confirmed the intrusion (late June 2026, via identity spoofing) but has not yet published its own victim count.
    • Exposed data for individuals includes name, address, household composition, reference tax income and withholding-tax rate.
    • The public portal impots.gouv.fr and personal accounts were NOT breached.
    • The real danger is phishing. France’s data-protection regulator (CNIL) and Paris prosecutors are now involved.

    First, the context: who is the DGFiP?

    The Direction générale des Finances publiques (DGFiP) is France’s national tax authority — it assesses and collects income tax, manages taxpayer records and runs the public tax portal impots.gouv.fr. Think of it as France’s IRS (US) or HMRC (UK). Two French fiscal concepts matter here. The revenu fiscal de référence (RFR), or “reference tax income,” is an official figure summarizing a household’s total income — it effectively reveals how much you earn and is used to set eligibility for many benefits. The prélèvement à la source is France’s pay-as-you-earn system; your personal withholding-tax rate also hints at your income level. Both were among the leaked fields — which is what makes this breach unusually sensitive.

    What happened

    On 12 August 2026, a hacker known as ZeroBytes claimed on a cybercrime forum to have exfiltrated 678,438 lines of data from the DGFiP. The next day, the French Ministry for Public Action and Accounts confirmed an “illegitimate access” to the agency’s information system, dating back to late June 2026.

    Notably, there was no software exploit or ransomware. The entry point was identity spoofing: someone impersonated an authorized user — using the credentials of an agent and/or an authorized third party — and queried an internal search tool. Access was cut off by late June during a routine control, but data had already been consulted and extracted.

    What data was exposed

    According to the breach-tracking platform FrenchBreaches, which reviewed samples of the files, individuals’ records include identity, date and place of birth, tax address, household situation and number of tax “parts,” plus the two sensitive figures above — reference tax income and withholding rate. Business records (company name, SIREN registration number, address) are considered less sensitive as they are often already public.

    The platform counts 678,437 people across the files (392,867 individuals and 285,570 businesses). Crucially, these numbers come from the attacker’s files, not from an official tally: the DGFiP says its investigation is ongoing. A second, unconfirmed claim on 14 August concerns cadastral (land-registry) data — potentially hundreds of thousands to over two million property rights-holders, depending on the source.

    Why it matters even outside France

    Anyone who has filed taxes in France — including foreign residents, cross-border workers and expatriates — could theoretically be in scope. The immediate threat is not stolen money but highly targeted phishing: with your name, address and real fiscal details, scammers can craft “tax refund” or “outstanding payment” emails that look authentic. Treat any unsolicited “impôts” or “tax office” message as suspicious, never click the link, and log in only by typing the official address yourself. The DGFiP never asks for bank details by email or text.

    The official response

    The DGFiP has apologized and says it will contact every affected person individually — by email or letter, starting the week of 18 August 2026 — specifying which data may have been accessed and what precautions to take. It has notified France’s data-protection authority, the CNIL, filed a criminal complaint, and the Paris public prosecutor’s cybercrime unit has opened an investigation.

    FAQ

    Was my data leaked?

    Only the DGFiP can confirm this. It will contact affected people directly from the week of 18 August 2026. The figures circulating come from the attacker’s files, not a final official count, so no one can confirm individual exposure from those numbers alone.

    What’s the actual risk?

    Mainly targeted phishing and identity fraud: criminals could use the exposed data to make scams convincing. No direct theft of bank funds has been reported through this incident.

    What should I do?

    Be wary of any unsolicited “tax office” email or text, never click embedded links, never share bank details, log in only via the official address you type yourself, and keep any official DGFiP notification you receive.

    Was impots.gouv.fr compromised?

    No. According to the administration, the public portal and users’ personal accounts were not breached. The intrusion targeted an internal tool via identity spoofing on the professional-access side.

    By Patrick Lancier

  • Free ChatGPT Is Now Unlimited — Here’s What Actually Changed (and What Didn’t)

    Free ChatGPT Is Now Unlimited — Here’s What Actually Changed (and What Didn’t)

    Free ChatGPT Is Now Unlimited — Here’s What Actually Changed (and What Didn’t)

    TL;DR — As of the week of August 10, 2026, free ChatGPT no longer has a message cap. OpenAI moved a recent model, GPT-5.6 “Luna” — previously reserved for paying subscribers (around $20–23/month) — into the free tier. You can now chat as much as you want, no credit card required. But “unlimited” doesn’t mean “everything is free.” Here’s the honest breakdown.

    What exactly changed

    On August 6, 2026, OpenAI announced it was removing the message limit on free ChatGPT. The old cap — roughly 10 messages, tightened earlier this year to 25 messages every 3 hours — is gone for text. In its place, the default model for Free and Go accounts becomes GPT-5.6 “Luna,” a newer, more reliable model. The rollout began the week of August 10 and is gradual, with no payment or even a credit card needed.

    What is GPT-5.6 “Luna”?

    Luna is the “fast and cheap” tier of OpenAI’s latest generation (the GPT-5.6 family, launched July 9, 2026). It’s built to answer quickly while costing little to run — which is precisely why OpenAI can afford to give it away without limits. On reliability, OpenAI claims 62–68% fewer factual errors than the previous generation. Paying subscribers keep access to a more capable model, GPT-5.6 “Sol,” which now powers both instant replies and deeper reasoning in one unified experience.

    The catch: “unlimited” applies to text

    This is the key nuance. The unlimited part covers text: asking questions, writing, summarizing, coding, translating, brainstorming. Several features stay limited or paid, however: file uploads, image generation and DALL-E 3. OpenAI has also added abuse guardrails and new safeguards for teens. So everyday chatting is effectively open-ended; heavier, creative, and professional workflows still favor the paid plans.

    Why now? A week that reshaped AI

    This didn’t happen in a vacuum. Between August 1 and 6, 2026, eight major AI models shipped from six different makers — DeepSeek, Alibaba, Black Forest Labs, Meta, OpenAI and xAI. Competition is fierce, and “free” has become the new battleground. By making its offering unlimited, OpenAI puts pressure on everyone else — and signals that access to a capable model is fast becoming a commodity, with the real value shifting to tools, integration, and volume.

    The best free alternatives

    ChatGPT isn’t your only option. Google Gemini (with its 2.0 Flash model) is widely seen as the most generous free tier. Mistral’s Le Chat offers a near-unlimited free version, hosted in France and GDPR-compliant by default — a strong pick if data privacy matters to you. Anthropic’s Claude remains a favorite for writing quality. And services like DeepSeek, HuggingChat and Pi impose no strict daily cap. Worth comparing based on your needs.

    Should you cancel your subscription?

    If you were paying only to escape the message cap, it’s worth checking whether the paid features — file uploads, image generation, professional workflows — actually matter to you. If they don’t, the free tier may now be enough. If they do, or if you rely on the more powerful Sol model, the subscription still earns its keep. Either way, free users just got a meaningfully better product for nothing.

    FAQ

    Is free ChatGPT really unlimited now?

    For text, yes: the message cap is gone, no credit card required. But file uploads, image generation and DALL-E 3 remain limited, and abuse guardrails apply.

    Which model do free users get?

    GPT-5.6 “Luna,” a recent, fast OpenAI model that used to be paid-only. Plus and Pro subscribers keep GPT-5.6 “Sol,” which is more capable.

    What are the best free alternatives?

    Google Gemini (very generous), Mistral Le Chat (near-unlimited, France-hosted, GDPR), Anthropic’s Claude (writing quality), plus DeepSeek, HuggingChat and Pi with no strict daily cap.

    Where’s the catch?

    “Unlimited” only covers text. Advanced features (files, images, DALL-E 3) stay limited or paid, and the offer is bounded by anti-abuse guardrails.


    By Patrick Lancier


  • Mark Walter Affair: Insurers Reclassify $21 Billion of Loans as Related-Party Transactions

    Mark Walter Affair: Insurers Reclassify $21 Billion of Loans as Related-Party Transactions


    NEWS · Insurance & Credit Markets

    Two life insurers in billionaire Mark Walter’s orbit — the Guggenheim Partners chief and Dodgers owner — reclassified roughly $21 billion of loans as related-party transactions after initially reporting $1 billion. Fitch now puts that exposure at nearly 40% of their portfolios. Manhattan federal prosecutors and the SEC are examining the disclosures. Here is what is established, and what is not.

    By Patrick Lancier — August 15, 2026

    TL;DR

    • The facts: Delaware Life Insurance Co. and Clear Spring Life and Annuity Co., both units of Group 1001 tied to Mark Walter, reclassified about $21 billion of loans as related-party transactions, versus $1 billion reported initially — including $4.6 billion at Clear Spring.
    • The diagnosis: per Fitch Ratings, these loans now reach ~40% of the insurers’ portfolios, up from about 2% — the highest such exposure among North American life insurers it rates.
    • The investigation: Manhattan federal prosecutors (SDNY) issued grand jury subpoenas in February 2026; a parallel SEC probe is under way. No charges have been filed.
    • The company’s response: Group 1001 says it is cooperating and that “our capital position and liquidity remain strong, and our financial strength ratings are unchanged.”
    • The liquidity lever: the record $12.5 billion Lakers sale (August 13, 2026) reportedly funds the reduction of affiliated assets.

    What the insurers reclassified

    At its core this is an accounting story before it is a criminal one. Delaware Life Insurance Company and Clear Spring Life and Annuity Company, both units of insurance and asset-management group Group 1001, had reported to regulators about $1 billion in loans to related parties. Following an internal review, they reclassified roughly $21 billion of loans as related-party transactions, including $4.6 billion held by Clear Spring, according to Fitch Ratings analyst Jamie Tucker. For Delaware Life alone, trade press reports affiliated holdings rising from roughly $1–1.4 billion to about $16–17 billion.

    The shift is not cosmetic. A life insurer’s assets back its obligations to policyholders and are subject to strict disclosure rules: a loan to an entity controlled by the same owner — here, companies tied to Mark Walter or his holding company TWG Global — must appear as an affiliated transaction, precisely because it can concentrate risk and blur the boundary between the insurer’s balance sheet and its owner’s ventures.

    Fitch’s read: from 2% to 40% of the portfolio

    It is the scale of the revision that drew market attention. According to Fitch, the reclassification lifts related-party loans from about 2% to nearly 40% of the two insurers’ portfolios — described by the agency as the highest such exposure among the North American life insurers it rates. From a prudential standpoint, a concentration of that magnitude raises questions of liquidity and governance: to what extent does policyholder money indirectly finance the group’s other activities?

    Why this is supervision, not (yet) sanction

    Precision matters here. This is neither a conviction nor proven fraud, but an accounting reclassification, a ratings-agency analysis and an ongoing investigation into the quality of regulatory disclosures. Such inquiries can close without any enforcement action. Every reading of the figures should therefore stay attributed: “per Fitch,” “according to the insurers’ regulatory filings,” “per Bloomberg.”

    The legal track: Manhattan and the SEC

    Delaware Life and Clear Spring received grand jury subpoenas in February 2026, disclosed in June 26 regulatory filings, according to Bloomberg. Manhattan prosecutors (Southern District of New York) are examining whether the insurers failed to disclose that their private-credit holdings backed other Walter-controlled vehicles; the inquiry runs in parallel with the SEC. The Financial Times first reported the SDNY probe, which several sources say was triggered by a whistleblower complaint concerning roughly $16 billion of private-credit deals. Notably, per Bloomberg, the subpoenas preceded the internal reviews that surfaced the reporting errors.

    The liquidity mechanics: Guggenheim pledged, Lakers cashed

    Two financial moves illuminate how the pressure is being managed. First, Walter offered to pledge his stake in Guggenheim Partners — the manager of about $320 billion in assets that he runs — as collateral for loans taken by TWG Global, with double-digit yields promised to lenders and a one-year term; if TWG fails to repay, creditors could seize and sell the collateral. Second, on August 13, 2026, Walter agreed to sell the Los Angeles Lakers to Josh Kushner and Bob Iger for a record $12.5 billion, little more than a year after acquiring control for about $10 billion. Per Bloomberg, the proceeds accelerate the cleanup of affiliated assets: Delaware Life and Clear Spring reportedly plan to report a reduction of up to $8 billion in affiliated assets in their next quarterly update.

    The company’s response

    Asked about the probe, Group 1001 said it is cooperating with investigators and that its financial condition remains strong: “Our capital position and liquidity remain strong, and our financial strength ratings are unchanged.” Neither Mark Walter, nor Guggenheim Partners, nor the insurers face charges as of publication.

    FAQ

    What is a related-party loan, and why does it matter for an insurer?

    Financing extended to an entity controlled by the same owner as the lender. For a life insurer, whose assets back policyholder obligations, such transactions must be disclosed separately because they can concentrate risk and create conflicts of interest.

    What are the key figures?

    Per Fitch, the insurers had reported ~$1 billion in related-party loans; after review they reclassified ~$21 billion, including $4.6 billion at Clear Spring. Fitch estimates these loans now reach ~40% of portfolios, up from ~2%.

    Has fraud been proven?

    No. As of mid-August 2026, no charges have been filed. The matter involves a Manhattan and SEC investigation into disclosure quality, plus an accounting reclassification and ratings analysis.

    How is this connected to the Lakers sale?

    Walter agreed on August 13, 2026 to sell the Lakers for $12.5 billion. Per Bloomberg, proceeds help reduce affiliated assets, expected to fall by up to $8 billion next quarter.

    References

    1. Bloomberg / Bloomberg Law, “Mark Walter’s Insurers, Guggenheim Probed by Prosecutors,” July 22, 2026. news.bloomberglaw.com
    2. Bloomberg, “Mark Walter Dangled Guggenheim Stake to Quickly Secure Loans,” August 13, 2026. bloomberg.com
    3. Financial Times (via Yahoo Finance), “Dodgers owner Mark Walter’s $16B in private-credit deals draws SEC and DOJ probes after whistleblower complaint,” 2026. finance.yahoo.com
    4. Insurance Business Magazine, “Mark Walter’s insurers face federal probe over undisclosed related-party investments,” 2026. insurancebusinessmag.com
    5. Yahoo Finance, “The obscure financial maneuver at issue in Dodgers owner probe explained,” 2026. finance.yahoo.com
    6. InvestmentNews, “Lakers sale spotlights Delaware Life’s $17B underreported exposure,” 2026. investmentnews.com
    7. Forbes, “The Los Angeles Lakers’ Record $12.5 Billion Sale Resets The Market For Sports Teams,” August 13, 2026. forbes.com
    8. Delaware Department of Insurance, “Examination Report of Clear Spring Life and Annuity Company as of 2023” (June 2025). insurance.delaware.gov
    9. Claims Journal, “Mark Walter’s Insurers, Guggenheim Probed by Prosecutors,” July 22, 2026. claimsjournal.com

    This is a financial news report. It relates regulatory filings, a ratings-agency analysis and ongoing investigations. No wrongdoing, illegality or fraud is established or attributed to any named person or company; all statements are attributed to their sources. Current as of August 15, 2026.

  • 800,000 Years Ago, Early Humans in Israel Already Planned How They Made Their Tools

    800,000 Years Ago, Early Humans in Israel Already Planned How They Made Their Tools

    TL;DR — A new study published in Scientific Reports finds that early humans at Gesher Benot Ya’aqov in northern Israel deliberately selected specific basalt sources for different stages of stone-tool production nearly 800,000 years ago. The finding points to advanced planning, deep environmental knowledge and traditions passed down across generations.

    Nearly 800,000 years ago, the early humans who lived on the shores of ancient Lake Hula were far from opportunistic toolmakers. According to a new study of stone tools from Gesher Benot Ya’aqov, they carefully chose which type of rock to use for each task — a level of foresight that reshapes how we think about early human intelligence.

    What did the researchers discover?

    The team found that the site’s inhabitants did not simply pick up whatever stone lay nearby. Instead, they deliberately sourced particular basalt flows for particular stages of tool-making. This selective behavior implies planning, memory of the landscape, and technological traditions that were maintained and repeated over long periods.

    Where is Gesher Benot Ya’aqov?

    Gesher Benot Ya’aqov (GBY) is an Acheulian archaeological site in northern Israel, on the shores of the former paleo-Lake Hula in the Jordan Valley. Dated to roughly 780,000 years ago, it preserves repeated occupations by early hominins and is one of the richest windows into life in the Middle Pleistocene.

    How did scientists trace the stone?

    Researchers used geochemical analysis to read the chemical “fingerprint” of each basalt artifact and match it to specific lava flows — including some that are now buried and invisible at the surface. By reconstructing a landscape that has changed dramatically over hundreds of thousands of years, they could show exactly where the toolmakers collected their raw material. The work was led by Dr. Tzahi Golan and Dr. Yoav Ben Dor of the Geological Survey of Israel, with Prof. Naama Goren-Inbar of the Hebrew University of Jerusalem.

    Why does this matter for human evolution?

    Deliberate raw-material selection is a hallmark of complex cognition. Showing that hominins were making these choices nearly 800,000 years ago pushes the evidence for long-term planning and cultural transmission much deeper into the human past than many models assumed.

    What is the Acheulian culture?

    The Acheulian is a prehistoric stone-tool tradition, best known for its teardrop-shaped hand-axes, that spanned roughly 1.7 million to 200,000 years ago. It is associated mainly with Homo erectus and related early humans, and Gesher Benot Ya’aqov is one of its most important sites outside Africa.

    FAQ — Early human tool-making at Gesher Benot Ya’aqov

    How old are the tools from Gesher Benot Ya’aqov?

    They date to about 780,000–800,000 years ago, during the Acheulian period of the Middle Pleistocene.

    What material were the tools made from?

    Mainly basalt, a volcanic rock, which the toolmakers sourced from specific lava flows around the site.

    Who carried out the study?

    Dr. Tzahi Golan and Dr. Yoav Ben Dor (Geological Survey of Israel) and Prof. Naama Goren-Inbar (Hebrew University of Jerusalem), published in Scientific Reports.

    Why is the discovery significant?

    It shows early humans deliberately selected raw materials, indicating advanced planning and cultural traditions far earlier than often assumed.

    Where is the site located?

    In northern Israel, on the shores of the former paleo-Lake Hula in the Jordan Valley.

    Sources

    By Patrick Lancier — published August 6, 2026.